Workbench capy git dialogue guardrails

Skill Fearvox/multica-ultimate-workbench/skills/workbench-capy-git-dialogue-guardrails

Public multi-agent workbench skill pack for SDD, review gates, context discipline, VM lanes, Sanity context, and autonomous agent coordination.

Install

npx -y skills add Fearvox/multica-ultimate-workbench --skill workbench-capy-git-dialogue-guardrails

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
23 stars23 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Source-first, self-loop resistant guardrails for Capy GitHub dialogue responders before any write-capable PR, issue, or review action.

SKILL.md

2.7 KB, as published. Nobody here has run it

Workbench Capy Git Dialogue Guardrails

Use this skill when implementing, reviewing, or dogfooding the Capy Git Dialogue Responder, especially for SYN-31, GitHub webhook responder work, Capy-authored PR/comment loops, or any task that could let Capy write back to a GitHub, Linear, or Slack surface.

This skill is source-layer only. It does not deploy Capy, enable webhooks, write Linear, post Slack, or mutate live OAuth/runtime state.

Read First

DECISIONS.md entry 2026-05-06 - Add Self-Loop Guardrails To The Capy Git Dialogue Responder
docs/capy-git-dialogue-responder.md
.capy/CAPTAIN.md, .capy/BUILD.md, and .capy/REVIEW.md when checking a repo-local Capy pack
autopilots/capy-linear-slack-sync.md only when Linear/Slack sync is in scope

Required Gate

Before any write-capable action, compute this block from source evidence:

CAPY_GIT_DIALOGUE_GUARDRAIL
event_author:
actor_classification:
human_request_present:
last_capy_commit_sha:
last_capy_comment_ids:
mutation_allowed:
circuit_breaker_state:
action_taken:
verdict: PASS | FLAG | BLOCK

Use the local dogfood helper for bounded fixture checks:

node scripts/capy-git-dialogue-guardrail.mjs --format json <event-summary.json>
node scripts/test-capy-git-dialogue-guardrail.mjs

Mutation Rules

Capy-authored comments, review comments, reviews, closeouts, commits, and synchronize events are observation only unless a human explicitly asks Capy to continue in that exact thread or after that exact commit.
The automatic patch budget is one patch attempt per PR per distinct human-authored review-finding batch.
Commit/comment/review churn from Capy itself is FLAG plus operator approval, not another patch loop.
Duplicate closeout for the same repo, PR, head SHA, and verdict tuple is a no-op.
Webhook receipt is not proof of repo state; verify PR, check, review, and git state from primary evidence.

Verdicts

PASS: mutation is allowed by a current human request and the patch budget is still available, or the action is read-only observation.
FLAG: mutation is refused because the trigger is self-authored, the patch budget is exhausted, or Capy churn requires operator approval.
BLOCK: the event summary is malformed or primary source identity cannot be established.

Closeout

CHANGED:
VERIFIED:
REMAINING:
PRS / LINKS:
VERDICT: PASS | FLAG | BLOCK

Do not claim live rollout. Source-layer PASS only means the guardrail source and dogfood fixtures passed.