abc/analysis_claude_code
1
0
Fork 0
mirror of https://github.com/shareAI-lab/analysis_claude_code.git synced 2026-09-23 22:03:41 +08:00
Code Issues Packages Projects Releases Wiki Activity

refactor: streamline the course to 17 lessons

Browse Source
This commit is contained in:
Haoran
2026-08-12 03:02:42 +08:00
parent ab35e59672
commit 7e2f2fd99b
250 changed files with 12179 additions and 18653 deletions
Download Patch File Download Diff File Expand all files Collapse all files

49
web/src/data/scenarios/s10.json
View File

@@ -1,27 +1,48 @@
{
"version": "s10",
"title": "Context Assembly",
"description": "Stable instructions and dynamic runtime state are assembled at the model boundary and cached by a deterministic context key.",
"title": "Task System",
"description": "A file-persisted task graph tracks status, ownership, and blockedBy dependencies.",
"steps": [
{
"type": "system_event",
"content": "context={workspace:\"/repo\", memories:true, skills:[\"code-review\"], tools:[\"bash\",\"read_file\"]}",
"annotation": "Prompt inputs are explicit runtime data."
"type": "user_message",
"content": "Break the release into tasks and block deployment until tests pass.",
"annotation": "The user asks for durable multi-step coordination."
},
{
"type": "system_event",
"content": "cache miss -> assemble_system_prompt(context)",
"annotation": "A new context key causes sections to be selected and joined."
"type": "tool_call",
"toolName": "create_task",
"content": "{\"subject\":\"Run tests\"}",
"annotation": "Tasks are created as JSON files under .tasks/."
},
{
"type": "assistant_text",
"content": "I can use the workspace, memory index, tool guidance, and available skills for this task.",
"annotation": "The LLM sees the assembled prompt, not a hardcoded static string."
"type": "tool_call",
"toolName": "create_task",
"content": "{\"subject\":\"Deploy release\",\"blockedBy\":[\"task_tests\"]}",
"annotation": "blockedBy encodes dependency ordering."
},
{
"type": "system_event",
"content": "same context key -> reuse cached prompt",
"annotation": "Repeated turns avoid rebuilding identical prompt text."
"type": "tool_call",
"toolName": "claim_task",
"content": "{\"task_id\":\"task_deploy\",\"owner\":\"agent\"}",
"annotation": "The claim fails until dependencies are complete."
},
{
"type": "tool_result",
"toolName": "claim_task",
"content": "Blocked by: [\"task_tests\"]",
"annotation": "The task graph prevents premature work."
},
{
"type": "tool_call",
"toolName": "complete_task",
"content": "{\"task_id\":\"task_tests\"}",
"annotation": "Completing a dependency can unblock downstream tasks."
},
{
"type": "tool_result",
"toolName": "complete_task",
"content": "Completed task_tests\nUnblocked: Deploy release",
"annotation": "The harness reports newly available work."
}
]
}

54
web/src/data/scenarios/s11.json
View File

@@ -1,33 +1,45 @@
{
"version": "s11",
"title": "Error Recovery",
"description": "LLM calls are wrapped with targeted recovery paths for token limits, prompt overflow, and transient provider errors.",
"title": "Background Tasks",
"description": "Long-running work moves to a thread while the agent keeps the main loop responsive.",
"steps": [
{
"type": "system_event",
"content": "try LLM call with max_tokens=8000",
"annotation": "The normal path is still a regular model call."
"type": "user_message",
"content": "Run a long repository scan and tell me when it finishes.",
"annotation": "The user asks for work that may exceed the normal foreground loop."
},
{
"type": "system_event",
"content": "error=max_tokens -> escalate to 64000 without appending a fake user message",
"annotation": "Token exhaustion gets a continuation-aware recovery path."
},
{
"type": "system_event",
"content": "error=prompt_too_long -> reactive_compact(messages) -> retry once",
"annotation": "Prompt overflow triggers compaction instead of blind retry."
},
{
"type": "system_event",
"content": "error=529 -> exponential backoff with jitter; repeated 529 -> fallback model",
"annotation": "Transient provider errors use backoff and model fallback."
"type": "assistant_text",
"content": "I'll start the scan in the background so we can keep the conversation responsive.",
"annotation": "The model chooses the background path instead of blocking on the tool result."
},
{
"type": "tool_call",
"toolName": "read_file",
"content": "{\"path\":\"README.md\"}",
"annotation": "After recovery succeeds, the agent continues the normal tool loop."
"toolName": "bash",
"content": "rg \"TODO|FIXME\" .",
"toolInput": "{\"background\":true}",
"annotation": "The tool call is wrapped with a background flag."
},
{
"type": "system_event",
"content": "background_tasks[task_42] = Thread(target=run_bash)",
"annotation": "The runtime starts a thread and returns immediately."
},
{
"type": "tool_result",
"toolName": "bash",
"content": "Started background task task_42.",
"annotation": "A placeholder result is appended to the main conversation."
},
{
"type": "system_event",
"content": "<task_notification id=\"task_42\" status=\"done\">Found 12 matches.</task_notification>",
"annotation": "The completed thread injects a notification back into the next loop."
},
{
"type": "assistant_text",
"content": "The scan finished with 12 matches. I'll summarize the most important ones now.",
"annotation": "The agent consumes the background result as fresh context."
}
]
}

56
web/src/data/scenarios/s12.json
View File

@@ -1,48 +1,44 @@
{
"version": "s12",
"title": "Task System",
"description": "A file-persisted task graph tracks status, ownership, and blockedBy dependencies.",
"title": "Cron Scheduler",
"description": "A daemon loop watches durable cron definitions, queues due jobs, and lets the agent process them.",
"steps": [
{
"type": "user_message",
"content": "Break the release into tasks and block deployment until tests pass.",
"annotation": "The user asks for durable multi-step coordination."
"content": "Every weekday morning, remind me to review open tasks.",
"annotation": "The user asks for a recurring future action."
},
{
"type": "assistant_text",
"content": "I'll create a cron job and store it so the scheduler can pick it up later.",
"annotation": "The agent delegates recurrence to the scheduler instead of relying on current context."
},
{
"type": "tool_call",
"toolName": "create_task",
"content": "{\"subject\":\"Run tests\"}",
"annotation": "Tasks are created as JSON files under .tasks/."
},
{
"type": "tool_call",
"toolName": "create_task",
"content": "{\"subject\":\"Deploy release\",\"blockedBy\":[\"task_tests\"]}",
"annotation": "blockedBy encodes dependency ordering."
},
{
"type": "tool_call",
"toolName": "claim_task",
"content": "{\"task_id\":\"task_deploy\",\"owner\":\"agent\"}",
"annotation": "The claim fails until dependencies are complete."
"toolName": "schedule_cron",
"content": "{\"name\":\"review-open-tasks\",\"cron\":\"0 9 * * 1-5\",\"prompt\":\"Review open tasks and summarize blockers.\"}",
"annotation": "The new tool validates and persists a cron definition."
},
{
"type": "tool_result",
"toolName": "claim_task",
"content": "Blocked by: [\"task_tests\"]",
"annotation": "The task graph prevents premature work."
"toolName": "schedule_cron",
"content": "Scheduled review-open-tasks with id cron_7.",
"annotation": "The job is written to the durable schedule store."
},
{
"type": "tool_call",
"toolName": "complete_task",
"content": "{\"task_id\":\"task_tests\"}",
"annotation": "Completing a dependency can unblock downstream tasks."
"type": "system_event",
"content": "cron_scheduler_loop: cron_7 matched current minute -> cron_queue.put(cron_7)",
"annotation": "The independent scheduler daemon detects that the job is due."
},
{
"type": "tool_result",
"toolName": "complete_task",
"content": "Completed task_tests\nUnblocked: Deploy release",
"annotation": "The harness reports newly available work."
"type": "system_event",
"content": "queue_processor_loop: dequeued cron_7 and invoked agent_loop(prompt)",
"annotation": "A separate processor turns due jobs into normal agent work."
},
{
"type": "assistant_text",
"content": "Scheduled. When the cron fires, the queue processor will run the reminder prompt through the agent loop.",
"annotation": "The final answer explains the durable recurring behavior."
}
]
}

121
web/src/data/scenarios/s13.json
View File

@@ -1,45 +1,122 @@
{
"version": "s13",
"title": "Background Tasks",
"description": "Long-running work moves to a thread while the agent keeps the main loop responsive.",
"title": "Agent Team Runtime",
"description": "The Lead coordinates persistent teammates while the runtime owns message delivery, atomic task claims, and task-bound worktree directories.",
"steps": [
{
"type": "user_message",
"content": "Run a long repository scan and tell me when it finishes.",
"annotation": "The user asks for work that may exceed the normal foreground loop."
"content": "Refactor authentication and update its dependent tests in parallel without mixing the edits.",
"annotation": "The user states the outcome and parallel-edit boundary, not an implementation-specific worktree procedure."
},
{
"type": "assistant_text",
"content": "I'll start the scan in the background so we can keep the conversation responsive.",
"annotation": "The model chooses the background path instead of blocking on the tool result."
"content": "I propose a backend teammate for authentication and a test teammate that picks up the dependent checks once they are ready. Shall I start the team?",
"annotation": "The Lead makes the extra-agent decision visible before spawning."
},
{
"type": "user_message",
"content": "Yes, go ahead.",
"annotation": "The team starts only after explicit confirmation."
},
{
"type": "tool_call",
"toolName": "bash",
"content": "rg \"TODO|FIXME\" .",
"toolInput": "{\"background\":true}",
"annotation": "The tool call is wrapped with a background flag."
},
{
"type": "system_event",
"content": "background_tasks[task_42] = Thread(target=run_bash)",
"annotation": "The runtime starts a thread and returns immediately."
"toolName": "create_task",
"content": "{\"subject\":\"Refactor authentication\"}",
"annotation": "The shared task board remains the source of ownership and status."
},
{
"type": "tool_result",
"toolName": "bash",
"content": "Started background task task_42.",
"annotation": "A placeholder result is appended to the main conversation."
"toolName": "create_task",
"content": "Created task_1a2b3c4d: Refactor authentication",
"annotation": "The runtime-generated task ID is carried into every later operation on this task."
},
{
"type": "tool_call",
"toolName": "create_worktree",
"content": "{\"name\":\"auth-refactor\",\"task_id\":\"task_1a2b3c4d\"}",
"annotation": "The worktree directory is recorded on the task instead of managed as a separate workflow or security sandbox."
},
{
"type": "tool_call",
"toolName": "create_task",
"content": "{\"subject\":\"Update authentication tests\",\"blockedBy\":[\"task_1a2b3c4d\"]}",
"annotation": "The task graph keeps dependent work from starting early."
},
{
"type": "tool_result",
"toolName": "create_task",
"content": "Created task_5e6f7a8b: Update authentication tests (blockedBy: task_1a2b3c4d)",
"annotation": "The second generated ID names the dependent task that the test teammate will later claim."
},
{
"type": "tool_call",
"toolName": "spawn_teammate",
"content": "{\"name\":\"backend\",\"role\":\"backend engineer\",\"prompt\":\"Review the assigned authentication task and propose a plan.\",\"task_id\":\"task_1a2b3c4d\",\"require_plan\":true}",
"annotation": "The runtime claims the initial task and activates the plan gate before the teammate thread starts."
},
{
"type": "tool_call",
"toolName": "spawn_teammate",
"content": "{\"name\":\"tests\",\"role\":\"test engineer\",\"prompt\":\"Watch the board and claim the dependent test task when it becomes ready.\"}",
"annotation": "A second persistent teammate can wait in IDLE without another direct dispatch."
},
{
"type": "system_event",
"content": "<task_notification id=\"task_42\" status=\"done\">Found 12 matches.</task_notification>",
"annotation": "The completed thread injects a notification back into the next loop."
"content": "spawn_teammate(backend, task_1a2b3c4d) -> task_store_lock commits owner=backend before thread start",
"annotation": "The ownership check and persisted state transition share the cross-process task-store lock."
},
{
"type": "tool_call",
"toolName": "request_plan",
"content": "{\"teammate\":\"backend\",\"task\":\"Inspect the claimed authentication task and submit a plan before changing files.\"}",
"annotation": "The Lead delivers the plan request for the current assignment; the gate was already active before the teammate thread started."
},
{
"type": "system_event",
"content": "backend submit_plan -> plan_approval_request(request_id=req_000007, task_id=task_1a2b3c4d)",
"annotation": "The request records the task and work version that the plan is meant to authorize."
},
{
"type": "tool_call",
"toolName": "review_plan",
"content": "{\"request_id\":\"req_000007\",\"approve\":true,\"feedback\":\"Proceed with the scoped refactor.\"}",
"annotation": "Approval is correlated by request ID and cannot carry into a different assignment."
},
{
"type": "system_event",
"content": "plan_approval_response(request_id=req_000007, approve=true) -> backend",
"annotation": "The teammate receives the typed response before mutating tools are released."
},
{
"type": "system_event",
"content": "backend tool cwd -> .worktrees/auth-refactor",
"annotation": "Bash, read, and write derive their directory from the claimed task binding."
},
{
"type": "tool_call",
"toolName": "complete_task",
"content": "{\"task_id\":\"task_1a2b3c4d\"}",
"annotation": "Completing the first task makes its dependent test task ready."
},
{
"type": "system_event",
"content": "backend -> Lead: result(auth refactor complete) -> idle_notification",
"annotation": "The task directory stays selected through the completion turn, then IDLE releases the assignment."
},
{
"type": "system_event",
"content": "claim_next_task(tests) -> task_5e6f7a8b; task_store_lock commits owner=tests",
"annotation": "An idle teammate discovers newly ready work without another direct assignment."
},
{
"type": "system_event",
"content": "tests -> Lead: result(test suite passed) -> idle_notification",
"annotation": "The runtime wakes the Lead when mailbox events arrive instead of asking the model to poll."
},
{
"type": "assistant_text",
"content": "The scan finished with 12 matches. I'll summarize the most important ones now.",
"annotation": "The agent consumes the background result as fresh context."
"content": "Authentication was refactored in its task-bound worktree and the dependent tests pass.",
"annotation": "The Lead combines parallel results into one user-facing outcome."
}
]
}

48
web/src/data/scenarios/s14.json
View File

@@ -1,44 +1,46 @@
{
"version": "s14",
"title": "Cron Scheduler",
"description": "A daemon loop watches durable cron definitions, queues due jobs, and lets the agent process them.",
"title": "MCP Tools",
"description": "The agent discovers external MCP tools and exposes them through a normalized tool namespace.",
"steps": [
{
"type": "user_message",
"content": "Every weekday morning, remind me to review open tasks.",
"annotation": "The user asks for a recurring future action."
},
{
"type": "assistant_text",
"content": "I'll create a cron job and store it so the scheduler can pick it up later.",
"annotation": "The agent delegates recurrence to the scheduler instead of relying on current context."
"content": "Search the documentation for deployment guidance.",
"annotation": "The user asks for a tool source outside the built-in set."
},
{
"type": "tool_call",
"toolName": "schedule_cron",
"content": "{\"name\":\"review-open-tasks\",\"cron\":\"0 9 * * 1-5\",\"prompt\":\"Review open tasks and summarize blockers.\"}",
"annotation": "The new tool validates and persists a cron definition."
"toolName": "connect_mcp",
"content": "{\"name\":\"docs\"}",
"annotation": "The runtime creates an MCP client for the named server."
},
{
"type": "system_event",
"content": "normalize_mcp_name(\"docs\", \"search\") -> mcp__docs__search",
"annotation": "External tools are namespaced to avoid collisions."
},
{
"type": "tool_result",
"toolName": "schedule_cron",
"content": "Scheduled review-open-tasks with id cron_7.",
"annotation": "The job is written to the durable schedule store."
"toolName": "connect_mcp",
"content": "Connected to MCP server 'docs'. Discovered 2 tools: search, get_version",
"annotation": "Tool discovery expands the active tool pool."
},
{
"type": "system_event",
"content": "cron_scheduler_loop: cron_7 matched current minute -> cron_queue.put(cron_7)",
"annotation": "The independent scheduler daemon detects that the job is due."
"type": "tool_call",
"toolName": "mcp__docs__search",
"content": "{\"query\":\"deployment\"}",
"annotation": "The LLM can now call the discovered tool by its normalized name."
},
{
"type": "system_event",
"content": "queue_processor_loop: dequeued cron_7 and invoked agent_loop(prompt)",
"annotation": "A separate processor turns due jobs into normal agent work."
"type": "tool_result",
"toolName": "mcp__docs__search",
"content": "[docs] Found 3 results for 'deployment'",
"annotation": "The external result is appended like any other tool result."
},
{
"type": "assistant_text",
"content": "Scheduled. When the cron fires, the queue processor will run the reminder prompt through the agent loop.",
"annotation": "The final answer explains the durable recurring behavior."
"content": "The docs server found three matches for deployment guidance.",
"annotation": "The agent summarizes external tool output for the user."
}
]
}

122
web/src/data/scenarios/s15.json
View File

@@ -1,122 +1,102 @@
{
"version": "s15",
"title": "Agent Team Runtime",
"description": "The Lead coordinates persistent teammates while the runtime owns message delivery, atomic task claims, and task-bound worktree directories.",
"title": "Integrated Harness",
"description": "The harness combines system instructions, tools, memory, teams, background work, cron, worktrees, and MCP.",
"steps": [
{
"type": "user_message",
"content": "Refactor authentication and update its dependent tests in parallel without mixing the edits.",
"annotation": "The user states the outcome and parallel-edit boundary, not an implementation-specific worktree procedure."
"content": "Prepare this project for release. Check the code, update the docs, and report deployment readiness.",
"annotation": "One ordinary request needs several earlier capabilities to work together."
},
{
"type": "system_event",
"content": "assemble_prompt: memory + task graph + skills + available tools + policy",
"annotation": "The runtime builds the prompt from layered sources of context."
},
{
"type": "assistant_text",
"content": "I propose a backend teammate for authentication and a test teammate that picks up the dependent checks once they are ready. Shall I start the team?",
"annotation": "The Lead makes the extra-agent decision visible before spawning."
"content": "I suggest a docs teammate while I run the release checks and inspect deployment status. Shall I start that team?",
"annotation": "The Lead proposes the team before adding another persistent agent."
},
{
"type": "user_message",
"content": "Yes, go ahead.",
"annotation": "The team starts only after explicit confirmation."
"content": "Yes. Start the docs teammate and continue.",
"annotation": "User confirmation opens the team boundary inherited from S13."
},
{
"type": "tool_call",
"toolName": "load_skill",
"content": "{\"name\":\"code-review\"}",
"annotation": "Skills contribute procedural context before execution."
},
{
"type": "tool_call",
"toolName": "create_task",
"content": "{\"subject\":\"Refactor authentication\"}",
"annotation": "The shared task board remains the source of ownership and status."
"content": "{\"subject\":\"Update release notes\",\"description\":\"Prepare release documentation and report the result.\"}",
"annotation": "The shared task board gives the docs work a stable owner and lifecycle."
},
{
"type": "tool_result",
"toolName": "create_task",
"content": "Created task_1712345678_0042: Refactor authentication",
"annotation": "The runtime-generated task ID is carried into every later operation on this task."
"content": "Created task_1712345678_0042: Update release notes",
"annotation": "The task ID connects the assignment to its working directory."
},
{
"type": "tool_call",
"toolName": "create_worktree",
"content": "{\"name\":\"auth-refactor\",\"task_id\":\"task_1712345678_0042\"}",
"annotation": "The worktree directory is recorded on the task instead of managed as a separate workflow or security sandbox."
},
{
"type": "tool_call",
"toolName": "create_task",
"content": "{\"subject\":\"Update authentication tests\",\"blockedBy\":[\"task_1712345678_0042\"]}",
"annotation": "The task graph keeps dependent work from starting early."
"content": "{\"name\":\"release-docs\",\"task_id\":\"task_1712345678_0042\"}",
"annotation": "The pending docs task receives a separate checkout before it is claimed."
},
{
"type": "tool_result",
"toolName": "create_task",
"content": "Created task_1712345678_0043: Update authentication tests (blockedBy: task_1712345678_0042)",
"annotation": "The second generated ID names the dependent task that the test teammate will later claim."
"toolName": "create_worktree",
"content": "Worktree 'release-docs' created for task_1712345678_0042",
"annotation": "The task now carries the checkout used by its eventual owner."
},
{
"type": "tool_call",
"toolName": "spawn_teammate",
"content": "{\"name\":\"backend\",\"role\":\"backend engineer\",\"prompt\":\"Claim the authentication task and propose a plan.\",\"require_plan\":true}",
"annotation": "The plan gate is active before the teammate thread starts, so claim and mutation cannot race ahead of approval."
"content": "{\"name\":\"docs\",\"role\":\"documentation engineer\",\"prompt\":\"Claim task_1712345678_0042, update the release notes in its task-bound worktree, and report the result.\"}",
"annotation": "Team delegation handles a parallel slice of the release."
},
{
"type": "tool_call",
"toolName": "spawn_teammate",
"content": "{\"name\":\"tests\",\"role\":\"test engineer\",\"prompt\":\"Watch the board and claim the dependent test task when it becomes ready.\"}",
"annotation": "A second persistent teammate can wait in IDLE without another direct dispatch."
"toolName": "bash",
"content": "{\"command\":\"python -m unittest tests.test_agent_teams_runtime\",\"run_in_background\":true}",
"annotation": "Long-running validation goes through the background task path."
},
{
"type": "system_event",
"content": "claim_next_task(backend) -> task_1712345678_0042; task_store_lock commits owner=backend",
"annotation": "The ownership check and persisted state transition share the cross-process task-store lock."
"content": "permission: user approved the exact test command",
"annotation": "Team confirmation does not authorize shell execution; the foreground turn asks separately before dispatch."
},
{
"type": "tool_call",
"toolName": "request_plan",
"content": "{\"teammate\":\"backend\",\"task\":\"Inspect the claimed authentication task and submit a plan before changing files.\"}",
"annotation": "The Lead delivers the plan request for the current assignment; the gate was already active before the teammate thread started."
},
{
"type": "system_event",
"content": "backend submit_plan -> plan_approval_request(request_id=req_000007, task_id=task_1712345678_0042)",
"annotation": "The request records the task and work version that the plan is meant to authorize."
"toolName": "connect_mcp",
"content": "{\"name\":\"deploy\"}",
"annotation": "External capabilities are added only when needed."
},
{
"type": "tool_call",
"toolName": "review_plan",
"content": "{\"request_id\":\"req_000007\",\"approve\":true,\"feedback\":\"Proceed with the scoped refactor.\"}",
"annotation": "Approval is correlated by request ID and cannot carry into a different assignment."
"toolName": "mcp__deploy__status",
"content": "{\"service\":\"web\"}",
"annotation": "Connecting discovers the tool; a separate call performs the status check."
},
{
"type": "tool_result",
"toolName": "mcp__deploy__status",
"content": "[deploy] web: running (v1.4.2)",
"annotation": "The readiness statement now comes from the MCP tool result rather than the connection event."
},
{
"type": "system_event",
"content": "plan_approval_response(request_id=req_000007, approve=true) -> backend",
"annotation": "The teammate receives the typed response before mutating tools are released."
},
{
"type": "system_event",
"content": "backend tool cwd -> .worktrees/auth-refactor",
"annotation": "Bash, read, and write derive their directory from the claimed task binding."
},
{
"type": "tool_call",
"toolName": "complete_task",
"content": "{\"task_id\":\"task_1712345678_0042\"}",
"annotation": "Completing the first task makes its dependent test task ready."
},
{
"type": "system_event",
"content": "backend -> Lead: result(auth refactor complete) -> idle_notification",
"annotation": "The task directory stays selected through the completion turn, then IDLE releases the assignment."
},
{
"type": "system_event",
"content": "claim_next_task(tests) -> task_1712345678_0043; task_lock commits owner=tests",
"annotation": "An idle teammate discovers newly ready work without another direct assignment."
},
{
"type": "system_event",
"content": "tests -> Lead: result(test suite passed) -> idle_notification",
"annotation": "The runtime wakes the Lead when mailbox events arrive instead of asking the model to poll."
"content": "task_notification(status=completed): tests passed; teammate result and deploy status appended",
"annotation": "The integrated runtime folds asynchronous results back into the loop."
},
{
"type": "assistant_text",
"content": "Authentication was refactored in its task-bound worktree and the dependent tests pass.",
"annotation": "The Lead combines parallel results into one user-facing outcome."
"content": "Release preparation is ready: checks passed, docs are updated in the teammate worktree, and deployment status is available.",
"annotation": "The integrated harness returns a single user-facing outcome."
}
]
}

62
web/src/data/scenarios/s16.json
View File

@@ -1,46 +1,54 @@
{
"version": "s16",
"title": "MCP Tools",
"description": "The agent discovers external MCP tools and exposes them through a normalized tool namespace.",
"title": "Workflow Runtime",
"description": "A fixed review pipeline executes inside one Workflow call, records a journal, emits lifecycle and progress events, and returns one final tool result.",
"steps": [
{
"type": "user_message",
"content": "Search the documentation for deployment guidance.",
"annotation": "The user asks for a tool source outside the built-in set."
"content": "Review this load_user change across correctness, security, performance, and style.",
"annotation": "The goal is open-ended, but the review procedure has a fixed shape."
},
{
"type": "tool_call",
"toolName": "connect_mcp",
"content": "{\"name\":\"docs\"}",
"annotation": "The runtime creates an MCP client for the named server."
"toolName": "Workflow",
"content": "{\"name\":\"review-changes\",\"args\":{\"budget\":null,\"changes\":\"def load_user(user_id):\\n query = f\\\"SELECT * FROM users WHERE id = {user_id}\\\"\\n return db.execute(query).fetchone()\\n\"}}",
"annotation": "The model selects a saved workflow and arguments; the host registry supplies its trusted metadata and script."
},
{
"type": "system_event",
"content": "normalize_mcp_name(\"docs\", \"search\") -> mcp__docs__search",
"annotation": "External tools are namespaced to avoid collisions."
"content": "async_launched(runId=wf_review-changes_0000000000001a7b) -> task_started",
"annotation": "The runtime emits launch lifecycle events before it executes the script; this is not a tool result."
},
{
"type": "system_event",
"content": "phase=Review -> pipeline([correctness, security, performance, style])",
"annotation": "Each item advances independently through the scripted stages."
},
{
"type": "system_event",
"content": "phase=Verify -> adversarial checks run in parallel",
"annotation": "Structured results cross a validation boundary before aggregation."
},
{
"type": "system_event",
"content": "workflow_agent status=done -> journal append",
"annotation": "Every completed agent call is checkpointed as the script runs."
},
{
"type": "system_event",
"content": "task_notification(status=completed, outputFile=.runtime/wf_review-changes_0000000000001a7b.output.json)",
"annotation": "The task emits its final lifecycle event after output is written."
},
{
"type": "tool_result",
"toolName": "connect_mcp",
"content": "Connected to MCP server 'docs'. Discovered 2 tools: search, get_version",
"annotation": "Tool discovery expands the active tool pool."
"toolName": "Workflow",
"content": "{\"launched\":{\"status\":\"async_launched\",\"taskId\":\"local_workflow_wf_review-changes_0000000000001a7b\",\"taskType\":\"local_workflow\",\"runId\":\"wf_review-changes_0000000000001a7b\",\"workflowName\":\"review-changes\"},\"result\":{\"confirmed\":[{\"dimension\":\"performance\",\"title\":\"audit:performance #1\",\"severity\":\"medium\"},{\"dimension\":\"performance\",\"title\":\"audit:performance #2\",\"severity\":\"medium\"},{\"dimension\":\"style\",\"title\":\"audit:style #1\",\"severity\":\"medium\"},{\"dimension\":\"style\",\"title\":\"audit:style #2\",\"severity\":\"medium\"},{\"dimension\":\"security\",\"title\":\"audit:security #1\",\"severity\":\"low\"},{\"dimension\":\"security\",\"title\":\"audit:security #2\",\"severity\":\"low\"}]},\"task\":{\"taskId\":\"local_workflow_wf_review-changes_0000000000001a7b\",\"taskType\":\"local_workflow\",\"runId\":\"wf_review-changes_0000000000001a7b\",\"workflowName\":\"review-changes\",\"status\":\"completed\",\"usage\":{\"agents\":11,\"tokens\":883},\"progress\":[{\"type\":\"workflow_phase\",\"title\":\"Review\"},{\"type\":\"workflow_agent\",\"label\":\"audit:correctness\",\"phase\":\"Review\",\"status\":\"done\"},{\"type\":\"workflow_phase\",\"title\":\"Verify\"},{\"type\":\"workflow_agent\",\"label\":\"audit:security\",\"phase\":\"Review\",\"status\":\"done\"},{\"type\":\"workflow_agent\",\"label\":\"audit:performance\",\"phase\":\"Review\",\"status\":\"done\"},{\"type\":\"workflow_agent\",\"label\":\"audit:style\",\"phase\":\"Review\",\"status\":\"done\"},{\"type\":\"workflow_agent\",\"label\":\"verify:correctness:audit:correctness #1\",\"phase\":\"Verify\",\"status\":\"done\"},{\"type\":\"workflow_agent\",\"label\":\"verify:security:audit:security #1\",\"phase\":\"Verify\",\"status\":\"done\"},{\"type\":\"workflow_agent\",\"label\":\"verify:security:audit:security #2\",\"phase\":\"Verify\",\"status\":\"done\"},{\"type\":\"workflow_agent\",\"label\":\"verify:performance:audit:performance #1\",\"phase\":\"Verify\",\"status\":\"done\"},{\"type\":\"workflow_agent\",\"label\":\"verify:performance:audit:performance #2\",\"phase\":\"Verify\",\"status\":\"done\"},{\"type\":\"workflow_agent\",\"label\":\"verify:style:audit:style #1\",\"phase\":\"Verify\",\"status\":\"done\"},{\"type\":\"workflow_agent\",\"label\":\"verify:style:audit:style #2\",\"phase\":\"Verify\",\"status\":\"done\"},{\"type\":\"workflow_log\",\"message\":\"confirmed 6 real finding(s)\"}]}}",
"annotation": "The demo fixture returns six synthetic findings and measured runner usage; these are not claims about the repository."
},
{
"type": "tool_call",
"toolName": "mcp__docs__search",
"content": "{\"query\":\"deployment\"}",
"annotation": "The LLM can now call the discovered tool by its normalized name."
},
{
"type": "tool_result",
"toolName": "mcp__docs__search",
"content": "[docs] Found 3 results for 'deployment'",
"annotation": "The external result is appended like any other tool result."
},
{
"type": "assistant_text",
"content": "The docs server found three matches for deployment guidance.",
"annotation": "The agent summarizes external tool output for the user."
"type": "system_event",
"content": "append Workflow tool_result -> messages[]",
"annotation": "A main-loop integration can append this JSON-safe result and continue with the updated conversation."
}
]
}

101
web/src/data/scenarios/s17.json
View File

@@ -1,102 +1,49 @@
{
"version": "s17",
"title": "Integrated Harness",
"description": "The harness composes context assembly, tools, memory, teams, background work, cron, worktrees, and MCP.",
"title": "Goal Loop",
"description": "A separate evaluator reads the conversation at the end of a turn and sends unfinished work back through the same loop.",
"steps": [
{
"type": "user_message",
"content": "Prepare this project for release. Check the code, update the docs, and report deployment readiness.",
"annotation": "One ordinary request needs several earlier capabilities to work together."
},
{
"type": "system_event",
"content": "assemble_prompt: memory + task graph + skills + available tools + policy",
"annotation": "The runtime builds the prompt from layered sources of context."
"content": "/goal pytest tests/auth exits with code 0",
"annotation": "The command stores an active completion condition and starts the work."
},
{
"type": "assistant_text",
"content": "I suggest a docs teammate while I run the release checks and inspect deployment status. Shall I start that team?",
"annotation": "The Lead proposes the team before adding another persistent agent."
"content": "The authentication code looks correct. The tests should pass now.",
"annotation": "With no more tool calls, the session asks the evaluator whether the active goal is complete."
},
{
"type": "user_message",
"content": "Yes. Start the docs teammate and continue.",
"annotation": "User confirmation opens the team boundary inherited from S15."
"type": "system_event",
"content": "goal_evaluated ok=false -> block",
"annotation": "The evaluator finds no test exit code in the conversation."
},
{
"type": "tool_call",
"toolName": "load_skill",
"content": "{\"name\":\"code-review\"}",
"annotation": "Skills contribute procedural context before execution."
},
{
"type": "tool_call",
"toolName": "create_task",
"content": "{\"subject\":\"Update release notes\",\"description\":\"Prepare release documentation and report the result.\"}",
"annotation": "The shared task board gives the docs work a stable owner and lifecycle."
},
{
"type": "tool_result",
"toolName": "create_task",
"content": "Created task_1712345678_0042: Update release notes",
"annotation": "The task ID connects the assignment to its working directory."
},
{
"type": "tool_call",
"toolName": "create_worktree",
"content": "{\"name\":\"release-docs\",\"task_id\":\"task_1712345678_0042\"}",
"annotation": "The pending docs task receives a separate checkout before it is claimed."
},
{
"type": "tool_result",
"toolName": "create_worktree",
"content": "Worktree 'release-docs' created for task_1712345678_0042",
"annotation": "The task now carries the checkout used by its eventual owner."
},
{
"type": "tool_call",
"toolName": "spawn_teammate",
"content": "{\"name\":\"docs\",\"role\":\"documentation engineer\",\"prompt\":\"Claim task_1712345678_0042, update the release notes in its task-bound worktree, and report the result.\"}",
"annotation": "Team delegation handles a parallel slice of the release."
"type": "system_event",
"content": "[Goal still active] Run pytest tests/auth and surface the exit code.",
"annotation": "The controller appends the reason to messages[] and continues the same loop; no queue is involved."
},
{
"type": "tool_call",
"toolName": "bash",
"content": "{\"command\":\"python -m unittest tests.test_agent_teams_runtime\",\"run_in_background\":true}",
"annotation": "Long-running validation goes through the background task path."
},
{
"type": "system_event",
"content": "permission: user approved the exact test command",
"annotation": "Team confirmation does not authorize shell execution; the foreground turn asks separately before dispatch."
},
{
"type": "tool_call",
"toolName": "connect_mcp",
"content": "{\"name\":\"deploy\"}",
"annotation": "External capabilities are added only when needed."
},
{
"type": "tool_call",
"toolName": "mcp__deploy__status",
"content": "{\"service\":\"web\"}",
"annotation": "Connecting discovers the tool; a separate call performs the status check."
"content": "pytest tests/auth",
"annotation": "The next agent turn runs the missing check."
},
{
"type": "tool_result",
"toolName": "mcp__deploy__status",
"content": "[deploy] web: running (v1.4.2)",
"annotation": "The readiness statement now comes from the MCP tool result rather than the connection event."
},
{
"type": "system_event",
"content": "task_notification(status=completed): tests passed; teammate result and deploy status appended",
"annotation": "The integrated runtime folds asynchronous results back into the loop."
"toolName": "bash",
"content": "exit_code=0\n12 passed in 0.42s",
"annotation": "The concrete command result enters the same conversation."
},
{
"type": "assistant_text",
"content": "Release preparation is ready: checks passed, docs are updated in the teammate worktree, and deployment status is available.",
"annotation": "The integrated harness returns a single user-facing outcome."
"content": "pytest tests/auth exited 0: 12 tests passed.",
"annotation": "A turn with no further tool call reaches the Stop hook again."
},
{
"type": "system_event",
"content": "goal_evaluated ok=true -> achieved",
"annotation": "The controller records success, clears the active goal, and returns."
}
]
}

54
web/src/data/scenarios/s18.json
View File

@@ -1,54 +0,0 @@
{
"version": "s18",
"title": "Workflow Runtime",
"description": "A fixed review pipeline executes inside one Workflow call, records a journal, emits lifecycle and progress events, and returns one final tool result.",
"steps": [
{
"type": "user_message",
"content": "Review this change across correctness, security, performance, and style.",
"annotation": "The goal is open-ended, but the review procedure has a fixed shape."
},
{
"type": "tool_call",
"toolName": "Workflow",
"content": "{\"name\":\"review-changes\",\"args\":{\"budget\":null}}",
"annotation": "The model selects a saved workflow and arguments; the host registry supplies its trusted metadata and script."
},
{
"type": "system_event",
"content": "async_launched(runId=wf_review-changes_0000000000001a7b) -> task_started",
"annotation": "The runtime emits launch lifecycle events before it executes the script; this is not a tool result."
},
{
"type": "system_event",
"content": "phase=Review -> pipeline([correctness, security, performance, style])",
"annotation": "Each item advances independently through the scripted stages."
},
{
"type": "system_event",
"content": "phase=Verify -> adversarial checks run in parallel",
"annotation": "Structured results cross a validation boundary before aggregation."
},
{
"type": "system_event",
"content": "workflow_agent status=done -> journal append",
"annotation": "Every completed agent call is checkpointed as the script runs."
},
{
"type": "system_event",
"content": "task_notification(status=completed, outputFile=.runtime/wf_review-changes_0000000000001a7b.output.json)",
"annotation": "The task emits its final lifecycle event after output is written."
},
{
"type": "tool_result",
"toolName": "Workflow",
"content": "{\"launched\":{\"status\":\"async_launched\",\"taskId\":\"local_workflow_wf_review-changes_0000000000001a7b\",\"taskType\":\"local_workflow\",\"runId\":\"wf_review-changes_0000000000001a7b\",\"workflowName\":\"review-changes\"},\"result\":{\"confirmed\":[{\"dimension\":\"security\",\"title\":\"audit:security #1\",\"severity\":\"high\"},{\"dimension\":\"style\",\"title\":\"audit:style #1\",\"severity\":\"high\"},{\"dimension\":\"security\",\"title\":\"audit:security #2\",\"severity\":\"medium\"},{\"dimension\":\"performance\",\"title\":\"audit:performance #2\",\"severity\":\"medium\"},{\"dimension\":\"correctness\",\"title\":\"audit:correctness #1\",\"severity\":\"low\"},{\"dimension\":\"performance\",\"title\":\"audit:performance #1\",\"severity\":\"low\"}]},\"task\":{\"taskId\":\"local_workflow_wf_review-changes_0000000000001a7b\",\"taskType\":\"local_workflow\",\"runId\":\"wf_review-changes_0000000000001a7b\",\"workflowName\":\"review-changes\",\"status\":\"completed\",\"usage\":{\"agents\":11,\"tokens\":352},\"progress\":[{\"type\":\"workflow_phase\",\"title\":\"Review\"},{\"type\":\"workflow_agent\",\"label\":\"audit:correctness\",\"phase\":\"Review\",\"status\":\"done\"},{\"type\":\"workflow_phase\",\"title\":\"Verify\"},{\"type\":\"workflow_agent\",\"label\":\"audit:security\",\"phase\":\"Review\",\"status\":\"done\"},{\"type\":\"workflow_agent\",\"label\":\"audit:performance\",\"phase\":\"Review\",\"status\":\"done\"},{\"type\":\"workflow_agent\",\"label\":\"audit:style\",\"phase\":\"Review\",\"status\":\"done\"},{\"type\":\"workflow_agent\",\"label\":\"verify:correctness:audit:correctness #1\",\"phase\":\"Verify\",\"status\":\"done\"},{\"type\":\"workflow_agent\",\"label\":\"verify:security:audit:security #1\",\"phase\":\"Verify\",\"status\":\"done\"},{\"type\":\"workflow_agent\",\"label\":\"verify:security:audit:security #2\",\"phase\":\"Verify\",\"status\":\"done\"},{\"type\":\"workflow_agent\",\"label\":\"verify:performance:audit:performance #1\",\"phase\":\"Verify\",\"status\":\"done\"},{\"type\":\"workflow_agent\",\"label\":\"verify:performance:audit:performance #2\",\"phase\":\"Verify\",\"status\":\"done\"},{\"type\":\"workflow_agent\",\"label\":\"verify:style:audit:style #1\",\"phase\":\"Verify\",\"status\":\"done\"},{\"type\":\"workflow_agent\",\"label\":\"verify:style:audit:style #2\",\"phase\":\"Verify\",\"status\":\"done\"},{\"type\":\"workflow_log\",\"message\":\"confirmed 6 real finding(s)\"}]}}",
"annotation": "The deterministic sample returns its six fixture findings and measured runner usage; these are not claims about the repository."
},
{
"type": "system_event",
"content": "append Workflow tool_result -> messages[]",
"annotation": "A main-loop integration can append this JSON-safe result and continue with the updated conversation."
}
]
}

49
web/src/data/scenarios/s19.json
View File

@@ -1,49 +0,0 @@
{
"version": "s19",
"title": "Goal Loop",
"description": "A separate evaluator reads the conversation at the end of a turn and sends unfinished work back through the same loop.",
"steps": [
{
"type": "user_message",
"content": "/goal pytest tests/auth exits with code 0",
"annotation": "The command stores an active completion condition and starts the work."
},
{
"type": "assistant_text",
"content": "The authentication code looks correct. The tests should pass now.",
"annotation": "With no more tool calls, the session asks the evaluator whether the active goal is complete."
},
{
"type": "system_event",
"content": "goal_evaluated ok=false -> block",
"annotation": "The evaluator finds no test exit code in the conversation."
},
{
"type": "system_event",
"content": "[Goal still active] Run pytest tests/auth and surface the exit code.",
"annotation": "The controller appends the reason to messages[] and continues the same loop; no queue is involved."
},
{
"type": "tool_call",
"toolName": "bash",
"content": "pytest tests/auth",
"annotation": "The next agent turn runs the missing check."
},
{
"type": "tool_result",
"toolName": "bash",
"content": "exit_code=0\n12 passed in 0.42s",
"annotation": "The concrete command result enters the same conversation."
},
{
"type": "assistant_text",
"content": "pytest tests/auth exited 0: 12 tests passed.",
"annotation": "A turn with no further tool call reaches the Stop hook again."
},
{
"type": "system_event",
"content": "goal_evaluated ok=true -> achieved",
"annotation": "The controller records success, clears the active goal, and returns."
}
]
}