feat(a2a): opt in to confidence-v1 + blast-v1 + hitl-mode-v1

[mabry1985]

Summary

Declares the three remaining workstacean x-protolabs extensions on Quinn's agent card, with full runtime emission for confidence-v1.

extension	shape	runtime work
confidence-v1	card + terminal DataPart	model self-reports via `` tags, parsed + emitted
blast-v1	card-only policy metadata	none
hitl-mode-v1	card-only approval policy	none

Design notes

confidence-v1 mirrors the cost-v1 pipeline:

1. `OUTPUT_FORMAT_INSTRUCTIONS` teaches the model to emit `0.85` + `<confidence_explanation>...` on the final turn.
2. `extract_confidence` pulls it out of the accumulated stream text, clamps to [0,1], returns (score, explanation).
3. `_chat_langgraph_stream` yields a `("confidence", {...})` event before `"done"`.
4. `a2a_handler._run_task_background` routes that into `_store.set_confidence`.
5. `_terminal_artifact_parts` appends a `CONFIDENCE_MIME` DataPart when set — `success` is derived from the terminal state so FAILED + high confidence correctly surfaces the calibration-warning signal on the workstacean side.
6. `_strip_reasoning` also drops `<confidence*>` tags so they never leak into user-facing output.

blast-v1 radii chosen to match actual skill behaviour — don't over-declare:

• `qa_report` → `self` (reads store, emits report)
• `board_audit` → `project` (scans one project's state)
• `pr_review` → `repo` (inspects a single repo's PRs)
• `bug_triage` → `project` (files a backlog feature on one board)

hitl-mode-v1: `qa_report` autonomous, the rest notification. None of Quinn's current skills warrant veto/gated/compound — pr_review leaves review comments but doesn't merge; bug_triage creates a reversible backlog feature.

Closes

#27

Test plan

• `tests/test_output_format.py` — extract_confidence (happy, absent, clamp, garbage, explanation-optional, doesn't leak into output)
• `tests/test_a2a_handler.py` — terminal artifact emits/omits confidence DataPart; success=false on FAILED; store.set_confidence clamps + no-ops on missing task
• `tests/test_a2a_integration.py` — card declares all three extensions; blast/hitl skills exist on the card; every radius/mode is one of the valid values
• 200/200 unit tests pass locally

[coderabbitai]

Warning

Rate limit exceeded

@mabry1985 has exceeded the limit for the number of commits that can be reviewed per hour. Please wait 0 minutes and 1 seconds before requesting another review.

Your organization is not enrolled in usage-based pricing. Contact your admin to enable usage-based pricing to continue reviews beyond the rate limit, or try again in 0 minutes and 1 seconds.

⌛ How to resolve this issue?

After the wait time has elapsed, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

We recommend that you space out your commits to avoid hitting the rate limit.

🚦 How do rate limits work?

CodeRabbit enforces hourly rate limits for each developer per organization.

Our paid plans have higher rate limits than the trial, open-source and free plans. In all cases, we re-allow further reviews after a brief timeout.

Please see our FAQ for further information.

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: ASSERTIVE

Plan: Pro

Run ID: 02b75f49-b86a-4320-a25d-21907ea28218

📥 Commits

Reviewing files that changed from the base of the PR and between 8da047f and 3c64cca.

📒 Files selected for processing (6)

• a2a_handler.py
• graph/output_format.py
• server.py
• tests/test_a2a_handler.py
• tests/test_a2a_integration.py
• tests/test_output_format.py

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch feat/ext-confidence-blast-hitl

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}