feat(skills): own tasks through verified outcomes

[0xr3ngar]

Rewrite the skills around completing and checking the requested work. Implementation tasks produce a draft PR, notify the user, and follow their review comments through fixes and resolved threads. Once the user marks the PR ready, the agent explains feedback and recommends what to address, then waits for their decision. Comments from other people or bots require the user's direction in either state.

Listening needs a running session or host support that wakes the agent. The skills describe that requirement, reviewer identity, and PR state checks; this PR does not add a background service.

Add bStack's philosophy with a short inspiration note crediting Theo (@t3dotgg), two quotes, and timestamped video links. Update the template and contributor guide, preserve functional comment directives, and leave uncertain time allocations unattributed. Include a minor Changeset.

All 30 existing tests, type checks, repository metadata checks, and standalone installations passed during the rewrite. Metadata and installation checks passed again after the review workflow edit. The instructions have not been evaluated in separate agent sessions. The bundled skill-creator validator rejects existing metadata extensions accepted by the repository validator and installer.

---

Note

Low Risk
Markdown-only changes to agent instructions and docs; no executable code paths, though installed skills will behave differently for users who adopt the new guidance.

Overview
Reframes bStack around verified outcomes instead of step-by-step workflows. Adds philosophy.md (problem-first autonomy, draft PR ownership, PR listeners before polling) and links it from the README and contributing guide.

Implementation skills (coding-best-practices, yeet) now spell out end-to-end delivery: open a draft PR, act on the requesting user’s comments while draft (fix, verify, resolve threads), then after “ready for review” explain feedback with recommendations and wait before changing code. Host PR event subscriptions are preferred over polling.

Other skills are tightened to match: coding-best-practices drops the long TypeScript example appendix in favor of concise style rules plus the same PR loop; no-comments keeps functional lint/suppression directives unless removal still passes checks; time-report leaves unclear ticket splits unattributed and tightens Jira logging/duplicate avoidance; unslop / bro are shorter outcome-focused rewrites.

Contributor surface: skill template uses Scope / Work / Completion; minor Changeset for the package.

^{Reviewed by Cursor Bugbot for commit 777b7b3. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

Risk: medium. Approved. Cursor Bugbot was not present on the current head after polling, remaining checks passed, and no approval policy requires human review.

  

_{Sent by Cursor Approval Agent: Pull Request Approver}

[cursor]

Cursor Bugbot has reviewed your changes using high effort and found 2 potential issues.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 777b7b3. Configure here.}

[cursor]

Requesting-user comments lack identity rules

Medium Severity

coding-best-practices now auto-follows GitHub comments attributable to the requesting user, but it never says how to identify that account or not to treat the PR author, gh login, or a collaborator as that person. The Coding install group does not include yeet, where those checks live, so draft-mode fixes can run on the wrong comments.

 

^{Reviewed by Cursor Bugbot for commit 777b7b3. Configure here.}

[cursor]

Existing non-draft PRs follow draft flow

Medium Severity

yeet now keeps working when a branch already has a PR and says to preserve that PR's status, but the next steps always announce that a draft is open and then enter the draft comment loop. A ready or closed PR is described as a draft, and user comments can be handled with draft auto-fix rules until a later recheck.

Additional Locations (1)

• skills/yeet/SKILL.md#L30-L39

 

^{Reviewed by Cursor Bugbot for commit 777b7b3. Configure here.}