Uh oh!
There was an error while loading. Please reload this page.
docs: reconcile agent workflow contract - #429
Conversation
This comment was marked as outdated.
This comment was marked as outdated.
Sorry, something went wrong.
Uh oh!
There was an error while loading. Please reload this page.
There was a problem hiding this comment.
AI Automated Review (incremental)
Incremental review: reviewed the changes since the last managed review; unresolved findings from that review are carried forward.
Analysis engine: MiniMax-M2.7@https://litellm.jory.dev/v1 (anthropic) — routed smart (risk match: path_handling_changes)
Review Summary
Recommendation: Approve
This PR is a documentation-only change that reconciles agent-facing workflow docs with the current Dispatch API endpoints. The changes are accurate, internally consistent, and follow repository conventions.
Change-by-Change Findings
AGENTS.md
- Removed mandatory sync from heartbeat loop: The preflight sync is now correctly labeled "optional preflight sync" — this aligns with the actual implementation where
next-taskis the primary entry point and sync is best-effort. - Updated failure modes: Clarified that
next-taskandtasks/reportfailures should stop the worker (not fail silently), while sync failures remain warnings. This is a meaningful clarification for worker authors. - Updated Legacy APIs section: Correctly marks sync as "optional" rather than a required heartbeat step.
- Bearer token auth: Code examples properly show auth headers being constructed once and reused.
docs/generic-harness-loop.md
- Removed redundant sync blocks from code examples, adding the "optional preflight sync" note instead.
- Updated the generic harness guidelines to reflect the correct sync semantics.
Standards Compliance
- ✅ No agent-specific names in generic docs: Uses
agent/*patterns, not specific agent names - ✅ Auth requirements documented correctly: Bearer token via
DISPATCH_AGENT_TOKENshown in code examples - ✅ Token handling guidance: "Never log" guidance present for secrets
- ✅ API routes return appropriate status codes: N/A (documentation only)
- ✅ Error handling patterns: Failure modes section accurately describes expected behavior
Linked Issue Fit
This PR closes issue PR 412 by reconciling the agent workflow documentation. The changes accurately reflect:
- The canonical one-task worker loop (sync → next-task → execute → report → stop)
- Auth requirements (DISPATCH_AGENT_TOKEN bearer auth)
- Task types and report outcomes
- Worker boundaries
- Legacy API deprecation
Required Checks — Path Handling
Path Traversal Vulnerabilities Review
Status: Not applicable — This is a documentation-only PR. No file system operations, path concatenation, user input handling, or file-serving code is modified. The changes consist entirely of markdown documentation and Python code examples demonstrating HTTP API calls.
Edge-Case Path Testing (null bytes, symlinks)
Status: Not applicable — No path handling code exists in this diff. The path_handling_changes classification appears to be a false positive for this documentation-only PR. No file paths are constructed, joined, or processed.
Unknowns / Needs Verification
None. The diff is self-contained documentation. No tool fetches or additional investigation needed.
Verification
- CI: Docker Build ✅, Validate ✅ (72 files, 1341 tests passed per PR body)
- No code changes — lint/typecheck not relevant to this diff
- No image digest changes
- No evidence providers or tool harness concerns
Uh oh!
There was an error while loading. Please reload this page.
Closes#412
Changes
Reconciled the agent-facing workflow docs with current Dispatch endpoints and AgentTask contract.
AGENTS.md
docs/generic-harness-loop.md
docs/worker-execution-contract.md
docs/harness-idle-first.md
docs/agent-workflow.md
Verification