skillsaw Issue

SkillDev tools

Use when skillsaw reports a likely false positive, misses an edge case, behaves incorrectly, or needs an RFE. Gather a safe report and, only with explicit user permission, create a feedback bundle or open a GitHub issue.

Available today. Use it from your connected AI after setup.

Connect ahel once, and every AI you use reads what you have installed.

Then ask your AI: use the skillsaw Issue skill

What this skill tells your AI

The instructions your AI receives, as published by stbenjam/skillsaw in skills/skillsaw-issue/SKILL.md and read by ahel’s review.

Use this skill when skillsaw itself needs a GitHub issue: a false positive, a missed violation, a crash, incorrect autofix, an edge case in rule detection, or a request for enhancement (RFE). Do not use it to report a problem in the repository being linted unless that problem demonstrates incorrect skillsaw behavior or a useful feature gap.

Gather safe evidence first

Collect the minimum evidence needed to explain the issue without sending anything externally:

  • skillsaw version and the exact command
  • affected rule ID, file type, and relevant configuration
  • expected behavior and actual behavior for a bug
  • user goal and proposed outcome for an RFE
  • a minimal reproducer or a short sanitized excerpt

Never copy credentials, private repository content, or complete diagnostic output into an issue draft. Prefer a minimal synthetic reproducer.

Require separate permission for each external action

Ask the user two explicit questions before taking either action:

  1. "May I run skillsaw feedback to create a local diagnostic ZIP?"
  2. "May I open an issue in stbenjam/skillsaw?"

Do not run skillsaw feedback without permission for the first action. Do not create a GitHub issue without permission for the second action. Permission for one action does not authorize the other. If permission is absent, present a sanitized draft and ask for the missing approval.

Create a feedback bundle only when approved

Run skillsaw feedback <repository-path> with a concise --message that describes the observed behavior. The command creates a local ZIP; it does not submit anything. Do not add --include unless the user explicitly identifies the file to share.

Review the ZIP before sharing it. skillsaw does not scan file contents for secrets: anything --include or --config names is copied in verbatim, so read those files yourself and remove or replace anything sensitive. Keep the bundle local if the user does not also approve opening an issue.

Open an issue only when approved

Search existing issues first to avoid duplicates. Then open an issue in stbenjam/skillsaw with:

  • a specific title naming the affected rule, behavior, or requested capability
  • the minimal reproducer and exact command for a bug
  • expected and actual results for a bug, or the user goal and proposed outcome for an RFE
  • skillsaw version and relevant configuration when applicable
  • the reviewed feedback ZIP only when the user approves attaching it

State which sensitive details were omitted. Link the issue back to the user and summarize what was shared. Do not change labels, milestones, assignees, or project state unless the user specifically requests it.

Signals

GitHub stars
66
Forks
15
Last commit
Sep 2026
Advanced
Catalog kind
skill
Gateway key
skillsaw-issue
Source
github.com/stbenjam/skillsaw