skillZs
★ LIVE SKILL TAGS ★
>>> LIVE SKILLS INDEX <<<
* OPEN SOURCE *
NO LOGIN, NO TRACKING
※ REAL INSTALL DATA ※
← back to all skills
adeonir/agent-skills65 installs

project-index

Generates project context and codebase documentation for AI agents. Creates `.agents/` directory with project overview and deep codebase analysis (architecture, conventions, testing, integrations, checklist, workflows, features when vertical slicing detected, review). Use when starting work on a project, onboarding to an existing codebase, generating project documentation for agents, mapping codebase patterns, or incrementally refreshing the index from recent git changes. Triggers: "initialize .agents", "setup project index", "index project", "map codebase", "analyze codebase", "project overview", "codebase summary", "onboarding to this repo", "integrate feedback", "sync knowledge", "refresh codebase", "sync codebase", "patch index". Not for feature work, Obsidian session notes, or dev tooling setup like prettier or eslint.

How do I install this agent skill?

npx skills add https://github.com/adeonir/agent-skills --skill project-index
view source ↗

Is this agent skill safe to install?

  • Gen Agent Trust Hubpass

    The project-index skill generates project documentation by analyzing local files, including source code and configuration manifests. It operates entirely within the local environment, using standard commands for file management, and contains no evidence of malicious behavior, data exfiltration, or remote code execution.

  • Socketpass

    No alerts

  • Snykfail

    Risk: HIGH · 1 issue

  • Runlayerwarn

    1/17 files flagged

  • ZeroLeakspass

    Score: 93/100 · 2 sections analyzed

What does this agent skill do?

Project Index

Generate project context and codebase documentation for AI agents.

Triggers

  • One-time setup ("initialize", "setup project index", "index project") → initialize.md
  • Project context ("overview", "project context", "refresh project") → overview.md
  • Codebase summary ("map codebase", "analyze codebase", "summary") → dispatch the 6 codebase refs in parallel via sub-agent fan-out, then run review.md with the outputs as context
  • Incremental refresh ("refresh codebase", "sync codebase", "patch index") → refresh.md
  • Architecture (single doc refresh) → architecture.md
  • Conventions (single doc refresh) → conventions.md
  • Testing (single doc refresh) → testing.md
  • Integrations (single doc refresh) → integrations.md
  • Checklist (single doc refresh) → checklist.md
  • Workflows (single doc refresh) → workflows.md
  • Features (single doc refresh, vertically sliced projects only) → features.md
  • Self-assessment (post fan-out) → review.md
  • Integrate feedback ("integrate feedback", "sync knowledge", "integrate discoveries") → integrate-feedback.md
  • Merge policy (loaded by any ref on re-run) → merge-policy.md

Workflow

initialize --> overview + (codebase fan-out + review when brownfield)
refresh    --> git diff range --> subset fan-out + review
                                     (fallback: full fan-out)

initialize is the entrypoint for first-time setup. It always runs overview. If the project is brownfield (has source code), it dispatches the codebase fan-out: 6 sub-agents in parallel (architecture, conventions, testing, integrations, checklist, workflows), plus a 7th (features) when vertical slicing is detected, then the main agent runs review with the fan-out outputs as context.

refresh is the cheaper re-run path once .agents/codebase/ exists. It diffs git from the last sync marker, routes changed paths to the sub-agents whose domain was touched, and dispatches only that subset. Falls back to the full fan-out when thresholds (file count, manifest churn, entry-point change) indicate a broad refactor.

Codebase Fan-Out

When mapping the full codebase (initialize on brownfield, or "summary" trigger), dispatch all 6 codebase refs as independent sub-agents in the same turn. A 7th sub-agent (features) is dispatched only when the project shows vertical slicing (see features.md Detection Gate).

Sub-agentReadsWritesGate
architectureEntry points, dir tree, layer boundaries.agents/codebase/architecture.mdalways
conventionsRepresentative source files.agents/codebase/conventions.mdalways
testingTest files, test config.agents/codebase/testing.mdalways
integrationsAPI clients, DB models, env files.agents/codebase/integrations.mdalways
checklistScripts (package.json / Makefile), pre-commit config.agents/codebase/checklist.mdalways
workflowsEntry points to traced data flows.agents/codebase/workflows.mdalways
featuresVertical slice dirs, route prefixes, co-located test/handler/view.agents/codebase/features.mdvertical slicing detected

Each sub-agent reads only what its domain needs. Outputs land on disk — sub-agents do not return findings through context.

After all sub-agents finish, the main agent runs review.md with all outputs as context. Review cannot be split across sub-agents — the main agent owns this synthesis.

Guidelines

  • Read actual code files to extract patterns, not just list them
  • Document conventions as observed, not as prescribed
  • Focus on stable patterns and interfaces, not volatile implementation details
  • Update existing docs when re-running (merge, never overwrite)
  • Be thorough in .agents/codebase/*.md — these load on demand, not always in context

Anti-Pattern: Source Boundary Leak

.agents/codebase/*.md captures only current observable state. Never read .artifacts/ (briefs, PRDs, design docs, epics, roadmaps) as a source of codebase facts. Forward-looking content ((planned), (TBD), milestone tags, feature IDs, "shipped through feature X") belongs to planning artifacts, not the codebase map. If a module, route, or dependency is described in .artifacts/ but absent from the filesystem right now, it does not exist.

Anti-Pattern: Convention by Dependency Name

Documenting conventions based on what's listed in package.json is guessing. Read the actual config files and source code to understand how the project uses (or extends, or overrides) each library. The project's actual values matter, not the framework's defaults.

Add the canonical catalog link to the repository README so users can inspect current installs and available audits. The publishing guide covers the complete discovery path.

<a href="https://skillzs.dev/skills/adeonir/agent-skills/project-index">View project-index on skillZs</a>