agentleFS
Sign inStart free
typvtypv/nest-turbo-starter/.claude/skills/fix/SKILL.md

fix

Skill224 starsChanged 3 months agoMITView on GitHub

ALWAYS activate this skill before fixing ANY bug, error, test failure, CI/CD issue, type error, lint, log error, UI issue, code problem.

We split this skill into the rules it gives an agent

6 rules. 0 are common, 1 is distinctive, and 5 are very specific to this project.

  • 0

    Common

    Most of the 69 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 5

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    First action: If there is no "auto" keyword in the request, use AskUserQuestion to determine workflow mode:

    line 23
  2. Distinctive:

    Create report with all findings for the next step.

    line 38
  3. Very specific:

    Task Orchestration (Moderate+ only): After classifying, create native Claude Tasks for all phases upfront with dependencies.

    line 51
  4. Very specific:

    Use TaskCreate with addBlockedBy for dependency chains

    line 53
  5. Very specific:

    Update via TaskUpdate as each phase completes

    line 54
  6. Very specific:

    Ask user if they want to commit via git-manager subagent

    line 68

“Common” is measured: kept by at least 6 in 10 of the 69 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Fixing

Unified skill for fixing issues of any complexity with intelligent routing.

Arguments

  • --auto - Activate autonomous mode (default)
  • --review - Activate human-in-the-loop review mode
  • --quick - Activate quick mode
  • --parallel - Activate parallel mode: route to parallel fullstack-developer agents per issue

Workflow

Step 1: Mode Selection

First action: If there is no "auto" keyword in the request, use AskUserQuestion to determine workflow mode:

OptionRecommend WhenBehavior
Autonomous (default)Simple/moderate issuesAuto-approve if score >= 9.5 & 0 critical
Human-in-the-loop ReviewCritical/production codePause for approval at each step
QuickType errors, lint, trivial bugsFast debug → fix → review cycle

See references/mode-selection.md for AskUserQuestion format.

Step 2: Debug

  • Activate debug skill.
  • Guess all possible root causes.
  • Spawn multiple Explore subagents in parallel to verify each hypothesis.
  • Create report with all findings for the next step.

Step 3: Complexity Assessment & Task Orchestration

Classify before routing. See references/complexity-assessment.md.

LevelIndicatorsWorkflow
SimpleSingle file, clear error, type/lintreferences/workflow-quick.md
ModerateMulti-file, root cause unclearreferences/workflow-standard.md
ComplexSystem-wide, architecture impactreferences/workflow-deep.md
Parallel2+ independent issues OR --parallel flagParallel fullstack-developer agents

Task Orchestration (Moderate+ only): After classifying, create native Claude Tasks for all phases upfront with dependencies. See references/task-orchestration.md.

  • Skip for Quick workflow (< 3 steps, overhead exceeds benefit)
  • Use TaskCreate with addBlockedBy for dependency chains
  • Update via TaskUpdate as each phase completes
  • For Parallel: create separate task trees per independent issue

Step 4: Fix Implementation & Verification

  • Implement fix per selected workflow, updating Tasks as phases complete.
  • Spawn multiple Explore subagents to verify no regressions.
  • Prevent future issues by adding comprehensive validation.

Step 5: Finalize (MANDATORY - never skip)

  1. Report summary: confidence score, changes, files
  2. docs-manager subagent → update ./docs if changes warrant (NON-OPTIONAL)
  3. TaskUpdate → mark ALL Claude Tasks completed
  4. Ask user if they want to commit via git-manager subagent

IMPORTANT: Skill/Subagent Activation Matrix

See references/skill-activation-matrix.md for complete matrix.

Always activate: debug (all workflows) Conditional: problem-solving, sequential-thinking, brainstorm-technical, brainstorm-features, context-engineering Subagents: debugger, researcher, planner, code-reviewer, tester, Bash Parallel: Multiple Explore agents for scouting, Bash agents for verification

Output Format

Unified step markers:

✓ Step 0: [Mode] selected - [Complexity] detected
✓ Step 1: Root cause identified - [summary]
✓ Step 2: Fix implemented - [N] files changed
✓ Step 3: Tests [X/X passed]
✓ Step 4: Review [score]/10 - [status]
✓ Step 5: Complete - [action taken]

References

Load as needed:

  • references/mode-selection.md - AskUserQuestion format for mode
  • references/complexity-assessment.md - Classification criteria
  • references/task-orchestration.md - Native Claude Task patterns for moderate+ workflows
  • references/workflow-quick.md - Quick: debug → fix → review
  • references/workflow-standard.md - Standard: full pipeline with Tasks
  • references/workflow-deep.md - Deep: research + brainstorm-technical + plan with Tasks
  • references/review-cycle.md - Review logic (autonomous vs HITL)
  • references/skill-activation-matrix.md - When to activate each skill
  • references/parallel-exploration.md - Parallel Explore/Bash/Task coordination patterns

Specialized Workflows:

  • references/workflow-ci.md - GitHub Actions/CI failures
  • references/workflow-logs.md - Application log analysis
  • references/workflow-test.md - Test suite failures
  • references/workflow-types.md - TypeScript type errors
  • references/workflow-ui.md - Visual/UI issues (requires design skills)

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

Reviews can't be read right now.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free