agentleFS
Data-WiseData-Wise/craft/skills/orchestration/task-analyzer/SKILL.md

task-analyzer

Skill4 starsChanged 12 days agoMITView on GitHub

This skill should be used when the user asks to "analyze a task", "route a command", "what craft command should I use", or describes a development task that needs routing to the appropriate craft workflow. Analyzes natural language task descriptions and maps them to optimal craft command sequences.

2 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 2 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

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

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

    When a task description reads like a fixed coded shape — decompose → cover N → verify M → synthesize — suggest /craft:orch:workflow, do not …

    line 88
  2. Distinctive:

    All category commands (code, test, arch, etc.)

    line 163
  3. Distinctive:

    All design skills (backend, frontend, devops)

    line 164

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “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

Task Analyzer Skill

Expert in analyzing natural language task descriptions and routing to appropriate craft commands.

When to Use

This skill is automatically activated when:

  • User invokes /craft:do <task>
  • User describes a development task
  • Context suggests multi-step workflow needed

Capabilities

Intent Recognition (D5, 2026-07-04: canonical source — /craft:do references

this table instead of duplicating it; see commands/do.md's Task Categories section)

Identifies task intent from natural language. Category names match /craft:do's routing categories exactly (renamed 2026-07-04 to eliminate a naming drift where this table used different labels — Create/Fix/Document/Refactor/Review — for the same 7 concepts /craft:do already routes on as Feature/Bug/Docs/Architecture, plus a missing Quality category /craft:do has and this table didn't):

IntentKeywordsExample
Featureadd, create, implement, build, new"add user authentication"
Bugfix, debug, resolve, repair, issue, error"fix the login bug"
Qualitylint, quality, clean, improve"clean up the utils module"
Testtest, verify, check, validate, coverage"test the api endpoints"
Docsdocument, docs, readme, changelog, explain"document the api"
Releaserelease, deploy, publish, ship"prepare for release"
Architecturedesign, refactor, restructure, review, audit, analyze"refactor the utils module"

Judgment call made here: the former standalone "Review" intent (review, audit, analyze) had no corresponding category or command mapping in /craft:do at all — folded its keywords into Architecture rather than inventing an 8th category /craft:do would have no routing for.

Domain Detection

Identifies affected domain from context:

DomainIndicatorsCommands
APIapi, endpoint, rest, graphqlarch:plan, code:test-gen
UIui, component, page, viewfrontend-designer skill
Databasedb, database, model, schemabackend-designer skill
Authauth, login, permissionarch:plan, code:test-gen
Testingtest, coverage, spectest, test --coverage
Docsdocs, readme, changelogdocs:sync, docs:validate

Workflow Selection

Maps intent + domain to optimal command sequence:

Feature Workflow:
  1. /craft:arch:plan     - Design the feature
  2. /craft:code:test-gen - Generate tests (TDD)
  3. dev/git skill        - Create feature branch (ask naturally; folded from /craft:git:branch)
  4. → Implementation     - User implements
  5. /craft:test          - Verify tests pass
  6. dev/git skill        - Commit/sync changes (ask naturally; folded from /craft:git:sync)

Bug Fix Workflow:
  1. /craft:code:debug    - Analyze the issue
  2. /craft:test debug    - Isolate in tests
  3. → Fix               - User implements fix
  4. /craft:test          - Verify fix
  5. dev/git skill        - Commit/sync changes (ask naturally; folded from /craft:git:sync)

Release Workflow:
  1. /craft:code:deps-audit  - Security scan
  2. /craft:test release     - Full test suite
  3. /craft:code:lint release - Comprehensive lint
  4. /craft:docs:changelog   - Update changelog
  5. /craft:code:release     - Release workflow

Coded-Workflow Shape Detection (D7)

When a task description reads like a fixed coded shape — decompose → cover N → verify M → synthesize — suggest /craft:orch:workflow, do not silently switch to it.

Detect the shape conservatively (the helper workflow_parse.detects_workflow_shape(text) fires only when ≥3 of the four stage categories appear, or the explicit decompose…synthesize chain is present):

Stage categoryTrigger words
decomposedecompose, break down, split into, dimensions, for each
fan-outfan out, in parallel, one per, cover each, per finding, reviewers
verifyverify, verifier, double-check, confirm each
synthesizesynthesize, aggregate, summarize, combine, merge findings

Suggest, never switch (routing-false-positive guard). A lone "verify" or "in parallel" must NOT route anywhere — auto-hijacking a task better served by improvised orchestrate is the accepted residual risk. On a match, present a suggestion the user confirms:

This looks like a fixed decompose → cover → verify → synthesize shape.
Consider: /craft:orch:workflow (coded, schema-gated, resumable)
Or keep going with improvised orchestration. Which do you want?
InputDetected?Action
"decompose into dimensions, review each in parallel, verify, synthesize"yessuggest :workflow
"fan out reviewers, verify findings, then summarize"yessuggest :workflow
"verify the login flow works"noroute normally
"review the architecture"noroute normally

Complexity Assessment

Determines task complexity for mode selection:

ComplexityIndicatorsDefault ModeMax Parallel Sessions
SimpleSingle file, quick fixdefault5
MediumFew files, featuredefault5
ComplexMany files, architecturedebugno cap
CriticalRelease, securityreleaseno cap

Advisory only — task-analyzer is used for command routing, not dispatch-time enforcement. Nothing in orchestrator-v2/craft:orch:workflow reads this table or blocks a dispatch that exceeds it (see #327).

Output Format

Analysis Report

╭─ Task Analysis ─────────────────────────────────────╮
│ Input: "add user authentication"                   │
├─────────────────────────────────────────────────────┤
│ Intent: Feature                                    │
│ Domain: Authentication                             │
│ Complexity: Medium                                 │
│ Workflow: Feature Development                      │
├─────────────────────────────────────────────────────┤
│ Recommended Commands:                              │
│   1. /craft:arch:plan       Design auth system    │
│   2. /craft:code:test-gen   Generate auth tests   │
│   3. dev/git skill          Create feature/auth   │
╰─────────────────────────────────────────────────────╯

Integration

Works with:

  • /craft:do - Primary entry point
  • /craft:check - Validation after workflows
  • All category commands (code, test, arch, etc.)
  • All design skills (backend, frontend, devops)

Example Analyses

Simple Task

Input: "fix typo in readme"
Analysis:
  - Intent: Bug
  - Domain: Documentation
  - Complexity: Simple
  - Command: Direct edit (no routing needed)

Complex Task

Input: "implement oauth2 login with google"
Analysis:
  - Intent: Feature
  - Domain: Authentication + API
  - Complexity: Complex
  - Workflow: Feature Development
  - Skills: backend-designer, devops-helper
  - Commands: arch:plan → code:test-gen → git:branch

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.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

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.