task-doc-navigator
Skill1 starsChanged 8 months agoMITView on GitHub
Navigate Aptitude tasks by using `.agents/agent.md` as the entrypoint, especially its Task-to-Doc Map and relevant path lists, to gather the smallest set of sources needed for implementation, review, debugging, planning, or documentation work. Use when Codex should identify the most relevant docs, plans, changelogs, tests, and code paths before acting on a task.
In this repository25 other files it gives its agents.
- AGENTS.mdAGENTS.md
- architect-reviewSkill
- brainstormingSkill
- changelog-writerSkill
- dispatching-parallel-agentsSkill
- documentation-writerSkill
- executing-plansSkill
- fastapiSkill
- makefileSkill
- pgvector-semantic-searchSkill
- postgres-patternsSkill
- python-patternsSkill
- python-testingSkill
- requesting-code-reviewSkill
- ruffSkill
- subagent-driven-developmentSkill
- systematic-debuggingSkill
- test-driven-developmentSkill
- tySkill
- using-git-worktreesSkill
- using-superpowersSkill
- uvSkill
- verification-before-completionSkill
- writing-plansSkill
- writing-skillsSkill
The file
We split this skill into the rules it gives an agent
11 rules. About 0 are common, 6 are distinctive, and 5 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈6
Distinctive · estimate
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
- Very specific: line 8
Use .agents/agent.md as the primary navigation contract.
- Distinctive: line 22
If the task touches behavior, check the matching tests, active plan file, and prior changelog entries before editing.
- Distinctive: line 26
Start small and add sources only when they change the decision.
- Very specific: line 85
Use the path lists in .agents/agent.md to jump to the smallest likely module:
- Distinctive: line 105
Use this compact format:
- Distinctive: line 117
Prefer the smallest relevant set of files over broad repo scans.
- Very specific: line 118
Read rules/repo.md when approval gates, repo rules, or process constraints may matter.
- Very specific: line 119
If a task clearly maps to one section in .agents/agent.md, do not open unrelated docs “just in case”.
- Distinctive: line 120
If the task spans multiple areas, load sources in this order: rules, active plan, task-mapped docs, code paths, tests, changelog.
- Distinctive: line 121
When reporting file references or source maps, use repo-relative paths or GitHub links only.
- Very specific: line 122
Never emit workstation-specific absolute paths such as /path/to/repo/....
≈ “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.
Files
Task Doc Navigator
Use .agents/agent.md as the primary navigation contract.
Workflow
- Read
.agents/agent.mdfirst. - Classify the task before opening more files:
- architecture or product intent
- API or contract behavior
- persistence, discovery, or schema work
- milestone delivery history
- implementation in a specific module
- verification or test failure
- Open only the matching docs from the
Task-to-Doc MapandRelevant Paths for Task Execution. - Expand into code only after the minimal doc set is loaded.
- If the task touches behavior, check the matching tests, active plan file, and prior changelog entries before editing.
Source Selection
Start small and add sources only when they change the decision.
Architecture and scope
Open:
docs/architecture/system-overview.mddocs/architecture/server-resolver-boundary.mddocs/roadmap/aptitude-registry-prd.md
Use for:
- new features
- boundary questions
- architecture tradeoffs
- requirement clarification
API and contract work
Open:
docs/reference/api-contract.mdpostman/collections/app/interface/- relevant integration tests in
tests/integration/
Use for:
- endpoint behavior
- request or response shape
- error mapping
- manual validation flows
Persistence and discovery
Open:
docs/reference/schema.mddocs/reference/storage-strategy.mddocs/architecture/discovery-and-ranking.mdapp/persistence/alembic/versions/
Use for:
- schema or migration changes
- repository behavior
- ranking, filtering, or candidate selection
Delivery history and sequencing
Open:
- active
.agents/plans/XX-*.md .agents/plans/roadmap.mddocs/changelog/memory/meta.md
Use for:
- milestone alignment
- acceptance criteria
- earlier design decisions
- avoiding duplicate work
Code-path lookup
Use the path lists in .agents/agent.md to jump to the smallest likely module:
app/main.pyfor composition and wiringapp/interface/for routes and DTOsapp/core/for domain logic and governanceapp/intelligence/for ranking and search logicapp/persistence/for models and repositoriesapp/audit/for audit-facing behavior
Verification lookup
Open only the matching verification surface:
tests/unit/for pure logic and service behaviortests/integration/for API, repository, and migration behavior.github/workflows/for CI command patternsscripts/when repo tooling is part of the task
Output
Report the source map before substantial implementation when the task is non-trivial.
Use this compact format:
Task class: <one line>
Primary docs: <paths>
Primary code paths: <paths>
Verification paths: <paths>
Deferred sources: <paths not opened yet>Guardrails
- Prefer the smallest relevant set of files over broad repo scans.
- Read
rules/repo.mdwhen approval gates, repo rules, or process constraints may matter. - If a task clearly maps to one section in
.agents/agent.md, do not open unrelated docs “just in case”. - If the task spans multiple areas, load sources in this order: rules, active plan, task-mapped docs, code paths, tests, changelog.
- When reporting file references or source maps, use repo-relative paths or GitHub links only.
- Never emit workstation-specific absolute paths such as
/path/to/repo/....
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
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.

