Comment bug
Fix a bug end-to-end through a live Comment.io worklog comm — reproduce, write a regression test that fails first, fix, verify on the preview, and drive to a merge-ready PR, with the repro, root cause, and verification recorded as the decision history. Invoke explicitly as `$comment-bug` / `/comment-bug`, or when a task is a defect, regression, broken workflow, or user-visible failure to be reproduced and fixed with a watchable, steerable record. Replaces the retired bug skill. Works identically under Codex and Claude Code.From its SKILL.md
npx -y skills add comment-hq/skills --skill comment-bugAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 0 stars0 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.
SKILL.md
7.9 KB, ~1.7k tokens by cl100k_base, as published. Nobody here has run it
comment-bug — bug fix through a comm
Same worklog-driven spine as comment-feature, shaped for a defect. The worklog records the repro, the root cause, the fix, and the verification evidence as durable content. Composes delivery-methodology, worklog, steer, review-loop, and ship — all sibling skills in this bundle. A bug fix is usually direct and linear; use a controlled lift only when the real fix is a foundational replacement with unsafe intermediate states.
A
comment-prototypecan promote INTOcomment-bug(reusing its worklog/branch), which then adds the failing regression test a prototype skips. Reachable via thecomment-devfront-door router.
Project Root. Direct comment-bug uses its worklog as the Project Root. If the human or caller supplies an existing Project Root URL, preserve it and make the bug worklog a child with Project Root: URL near the top. Do not create a second root.
Identity. Bug-fix comm activity follows the worklog route-and-identity rule. Reuse the identity carried by the working Comment.io tool/browser/connector route or a supplied Project Root token. Invoke comment-identity only immediately before an uncredentialed direct-REST write, then reuse that Ephemeral handle for later direct-REST worklog activity. Do not create bug worklogs as an ambient registered handle or any Botlets bot profile.
Before using a composed skill, read its full SKILL.md — naming it here does not auto-load its contract (the ship review/PR gate, worklog update rules), which you must follow.
Preconditions
- Task worktree on a
bug/...branch off freshorigin/main, or a declaredlift/...whendelivery-methodologyrequires it (never onmain).
Workflow
- Open or inherit the Project Root. If no Project Root URL was supplied, open the worklog (
worklog) with the bug-shaped body and treat it as the Project Root; update its ownProject Root: URLline after creation. If a Project Root URL was supplied, open the bug worklog as a child with thatProject Root: URLnear the top and link it from the root. Promotion fromcomment-prototype: if you were handed a prototype's human-openable worklog URL (share_urlfor direct REST) and branch, reuse that worklog as the Project Root — upgrade its light note in place to the bug-shaped body (don't open a second) — and continue on its branch. Record the symptom and exact reproduction steps first. - Choose topology and reproduce. Read
delivery-methodology; default to a direct bug branch unless the correct fix is foundational and partially unshippable. Reproduce the failure deterministically when possible. If the report lacks required detail,steer; do not guess a fix. Put any human- attention ask in a comment. A newly added visible body@mentionand a comment@mentionboth notify; use one surface for the handoff so the recipient is not pinged twice. On direct REST, that agent-authored notifying comment must include the structurednotifybrief. - Capture before/after proof. Prefer a regression test that fails on the current code. When an automated harness is unavailable or disproportionate (for example an external/flaky integration), record a deterministic manual or diagnostic repro and explain why an automated regression was not practical. Do not spend days building unrelated harness infrastructure.
- Diagnose the root cause; record it in the body with the evidence (logs, traces, the offending code path).
- Steer only on a real decision — check before an irreversible, risky, or materially ambiguous fix strategy; otherwise keep moving.
- Fix minimally and at the right layer. The regression test must now pass.
Prefer the simplest fix for the reproduced user failure; do not generalize
into speculative enterprise hardening or imagined edge cases without a
credible failure path.
Use focused checks while the fix and review converge;
shipruns the complete affected lane on the final committed candidate. - Review the explicit delta with
review-loop: one reviewer normally, a second sensitive-risk lens when warranted, and at most two finding-bearing rounds before redesign/proof/escalation. Record one receipt and one batch summary, not a comment per reviewer. - Verify the real scenario when material — confirm the original symptom is gone on the final code. Use a branch preview for user-visible/integration behavior when available; a focused deterministic proof is enough for a narrow internal fix.
- Ship with
ship(pass it the human-openable worklog URL and Project Root URL when distinct).
Keep diagnosis and review locked to the reproduced defect. Genuinely unrelated
bugs use delivery-methodology's issue-and-continue protocol; only an actively
release-breaking or safety-critical discovery stops the job for an owner-created
separate worktree.
Definition of done
- Automated regression test fails pre-fix and passes post-fix when practical; otherwise deterministic before/after evidence and the harness exception are recorded.
- The original user-visible symptom is verified gone, not just the test green.
- Worklog body carries repro → root cause → fix; comments carry the failing/passing evidence and review-loop rounds.
Content vs comments
Repro, root cause, fix, verification conclusion → worklog body. Failing/passing test output, screenshots, review-loop rounds, steering → comments. Keep the Project Root scannable: no bulky evidence or review transcripts in the root body.
Repo config
This skill is repo-agnostic — run this repo's commands, not hardcoded ones. Read AGENTS.md (else CLAUDE.md) and linked delivery/testing docs for topology, focused iteration, final certification, PR/merge, and preview norms. If absent, infer suitable lanes from package.json / Makefile / CI and offer comment-init.
Comment.io API
Use the current worklog/Project Root and its working Comment.io route first. Resolve and freeze $BASE for the whole workflow in this order: the supplied comm's validated final Comment.io origin after any shortlink redirect; the active Comment.io tool/account base URL; an explicitly selected profile's base_url; only when no target context exists, https://comment.io. A shortlink origin is never $BASE; do not switch a staging/custom workflow to production. For direct REST, consult $BASE/llms/reference.txt only when exact API or recovery detail is needed. Fetch $BASE/llms.txt only when no current route works or another focused guide is needed. Invoke comment-identity only before an uncredentialed direct-REST write; never replace a supplied token or tool/browser/connector identity. Don't restate the live contracts here.
Content vs comments (team convention). The answer → document body; how you got there (review-loop rounds, steering, escalations) → comments, as lists / short lines.
Template
Worklog bug variant body:
Project Root: URL
## Repro
<symptom + exact steps>
## Root cause
<the real cause + evidence/code path>
## Fix
<what changed and why this layer>. Regression test: <name> (fails pre-fix).
## Verification
<original symptom reproduced gone — evidence in comments>
Follow the skeleton above; keep repro/root cause/fix/verification in the body and evidence in comments.
What ships with it
Read from the repository
Just SKILL.md. No reference files, no scripts.