Index of every page on this site: llms.txt
All skills
Run it yourself

Use when given a bug report, failing test, stack trace, or CI failure that needs diagnosis and repair. Autonomously diagnoses root cause, implements minimal fix, verifies correctness, and commits. Keywords: fix bug, failing test, stack trace, CI failure, debug error, broken test. Triggers on "fix bug" or "debug failing test".

Use this Skill: https://skilld.dev/gh/acedergren/agentic-tools/bugfix

Nothing lands on disk. Nothing to clean up.

Fork this Skill

Edit a local copy. It keeps the author and licence.

SKILL.md

≈84 tokens for metadata: the name and description. ≈773 when used: this file.

Description uses 4.2% of example budget

Before choosing a Skill, your Agent reads its name and description. All available Skills share that space.

  • A shorter description leaves more room for other Skills. This entry exceeds our 1% size suggestion.

In our Claude Code example, all Skill names and descriptions share 8,000 characters. This Skill uses ≈336 characters, or 4.2%.

The 1% threshold is a size suggestion. Longer descriptions can still fit.

Your model, settings, and other Skills decide how much text your Agent can read.

Example settings and source

The example uses a 200k-token context and default Claude Code settings. The count includes the name, description, separators, and when_to_use when present. Codex also counts local file paths.

Skit's source and limits: Codex 0.160.1, Claude Code 2.1.292.

Bugfix

When to Use

Load this skill when the user request matches the frontmatter description for Bugfix.

Autonomous end-to-end bug fix: diagnose from evidence, find root cause, minimal fix, verify, commit.

NEVER

  • Never ask the user for more information unless the bug description is completely ambiguous — use tools to find it.
  • Never refactor unrelated code while fixing a bug — scope creep breaks regression isolation.
  • Never fix other bugs you notice while fixing the target — create a new task instead.
  • Never git add -A — stage only the files you changed.
  • Never skip running the full workspace test suite after a fix — targeted test passing doesn't mean no regression.
  • Never commit until typecheck passes on the affected workspace.

Root Cause Classification (decide before writing any code)

Type Signal Fix approach
Code bug Logic error, wrong query, bad comparison Fix implementation
Test bug Expectation contradicts documented contract Fix the test — explain why it was wrong
Mock wiring mockReset: true pattern not followed, stale mock state Use correct mock pattern from CLAUDE.md
Type error TypeScript compilation failure Fix types, not tests
Import error Wrong package path, missing export Fix import resolution

If it's a mock wiring issue: read the mock patterns in CLAUDE.md before writing anything — there are three distinct patterns (forwarding, object-bag, counter-based sequencing).

Evidence-First Thinking

Collect before hypothesizing:

  1. Run the failing test — see the exact error output, not just the description
  2. Read the failing test — understand what it expects
  3. Read the source under test
  4. git log --oneline -10 -- <file> — was this recently changed?
  5. For runtime errors: check logs, route/plugin/service code

State hypothesis explicitly before writing code: "The test fails because X, caused by Y."

Abort Conditions (stop and ask user)

  • Bug is in security-sensitive area (auth, RBAC, IDOR) — confirm fix approach first
  • Fix requires changing more than 5 files — may be an architectural issue
  • Root cause requires changing the mock strategy for an entire test suite

Scripts

bash scripts/collect-bugfix-context.sh apps/api/src/tests/routes/auth.test.ts
bash scripts/run-targeted-test.sh apps/api/src/tests/routes/auth.test.ts
bash scripts/run-targeted-test.sh apps/api/src/tests/routes/auth.test.ts "returns 500"

Verification Sequence

  1. Run the specific failing test — must pass
  2. Run the full workspace suite — no new failures
  3. Run typecheck on the affected workspace (npx tsc --noEmit)
  4. git diff --name-only — verify every changed file relates to the bug; revert anything that doesn't

Commit Format

fix(scope): what was broken and how it was fixed

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Arguments

$ARGUMENTS: Error message, failing test path, CI log URL, or reproduction steps. If empty, ask for the error or failing test name.

Source: SKILL.md on GitHub

skilld matched fixed text patterns in SKILL.md and file names. Patterns miss obfuscated code.

skilld run checks every file with the same patterns. It asks for approval before it loads a Skill with a behavior marked Needs approval.

1 warning7mo5 checks · Risk SAFE
  • Gen Agent Trust Hub8mo

    This skill provides an autonomous workflow for diagnosing and fixing software bugs using tools like Vitest and Git. The primary security risk is indirect prompt injection, where malicious instructions hidden in a bug report could influence the agent's actions. The skill includes reasonable safeguards, such as stopping for manual review on security-sensitive code paths.

  • Socket8mo

    No alerts

  • Snyk8mo

    Risk: LOW · No issues

  • Runlayer8mo

    3/3 files flagged

  • ZeroLeaks7mo

    Score: 93/100 · 2 sections analyzed

Signed by skilld at 641a79b. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub last month.

Steadyupdated 6 months ago

README badge

README badge for acedergren/agentic-tools/bugfix