fix(review-transition): report no-PR Workload verdict to dispatch instead of silent skip:no-pr

[itsmiso-ai]

The coder's branch correctly implements issue #213: no-PR routing now depends on the coder task's verdict — INCOMPLETE/absent verdicts park the issue as "blocked", while GO verdicts surface as anomalies with commitSHA/branch info without parking. …

Fixes #213

Opened by foreman on review GO (workload wl-misospace-foreman-dispatch-bridge-213).

[its-saffron]

AI Automated Review

Full PR review.

Analysis engine: MiniMax-M2.7@https://litellm.jory.dev/v1 (anthropic) — escalated (fast_low_confidence)

PR PR 214 Review — Approve

Recommendation: Approve. The PR correctly implements issue PR 213 and all required checks have been addressed.

---

Change-by-Change Findings

1. _extract_task_signal() (lines 51–74)

New helper function that reads verdict, result.summary, and result.extra from tasks. It iterates through tasks and returns the signal from the last task carrying a verdict (or "" for absent verdict). The implementation safely uses .get() with fallbacks and str() coercion — no unsafe type handling.

2. No-PR routing logic (lines 139–171)

The old results.append(f"{name}:skip:no-pr") is replaced with a two-branch conditional:

• Non-GO or absent verdict → "blocked": Calls update_status(item, "blocked", agent_name), matching the pattern referenced in bridge/claim.py:333. Results include the verdict value and optional summary.
• GO verdict with no PR → anomaly: Appends anomaly:go-no-pr with commitSHA, branch, and summary from extra. Does not call update_status (issue not parked).

Both branches include except Exception as e handlers matching the existing error pattern in the same loop.

---

Required Checks — Explicitly Addressed

1. Review for path traversal vulnerabilities

Status: Does not apply.

This PR operates entirely on in-memory data structures (dicts, strings) read from Workload/task manifests. It does not:

• Open, read, write, or link to filesystem paths
• Accept user-supplied paths
• Interpolate strings into paths, URLs, or command invocations

The commitSHA and branch values read from extra are only string-interpolated into result log lines (e.g. commitSHA=abc123), not into any path construction. No path traversal surface exists in this diff.

2. Test with edge-case paths (null bytes, symlinks)

Status: Does not apply.

The code under test is pure business logic (transition_to_in_review) that processes task verdict data in memory. It does not interact with the filesystem, create or follow symlinks, or parse path strings. Null-byte injection and symlink attacks are not relevant attack vectors for this change.

---

Standards Compliance

The AGENTS.md conventions for this repository govern issue filing (imperative ask + named file paths). This PR is a code implementation, not an issue, so those conventions do not apply to the diff. Code-level conventions observed:

• Docstrings explain behavior and rationale (e.g. _extract_task_signal)
• except Exception as e handling matches the existing pattern in the same loop (results.append(f"{name}:error:{e}"))
• Error messages are informative (API down in test)
• No new status values invented; "blocked" is confirmed via git_grep as an existing dispatch status

---

Linked Issue PR 213 Fit

Acceptance Criterion	Status
No-PR non-AllAlreadyResolved reports to dispatch (not silent)	✅ _extract_task_signal + result appends
Non-GO verdict parks as "blocked" (not "done")	✅ update_status(item, "blocked", ...)
GO verdict with no PR is distinct anomaly with commitSHA/branch	✅ anomaly:go-no-pr branch
status.result.summary carried into result line	✅ Appended via :{summary}
Fails open: raising update_status returns error line, not abort	✅ except Exception as e branch
Absent verdict treated as non-GO and parks	✅ verdict != "GO" catches empty string
AllAlreadyResolved still goes to done	✅ Test: test_all_already_resolved_still_goes_to_done

All six acceptance criteria are satisfied. The diff also passes the explicit non-regression: AllAlreadyResolved with a verdict present still goes to done.

---

Tool Harness Findings

Not applicable — no tool harness output in corpus.

---

Unknowns / Needs Verification

None. The diff is self-contained, tests are comprehensive, CI passed (test + docker: success), and the implementation matches the issue's routing logic specification exactly.