ahtishamshahzad/agent_dev_flow/documentation

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

Use to write or update documentation of the work inside `.ai/` — decisions, project state, work items, knowledge, generated reports — concisely and in its canonical place. For the applications' own documentation (the `docs/` tree), use `application-documentation` instead.

Install

npx vskill@latest install ahtishamshahzad/agent_dev_flow/documentation

SKILL.md

Documentation

Purpose

Produce and maintain documentation of the work — placed canonically, kept concise, and never duplicated. Implements ../../system/DOCUMENTATION_RULES.md.

Scope: everything under .ai/. The applications' own documentation — the docs/ tree with a folder per app and a file per screen, page, endpoint, or job — belongs to application-documentation. Both are needed; neither restates the other.

When to Use

  • Recording decisions, project state, work items, knowledge, or reports.
  • Updating READMEs/indexes when structure changes.
  • Not for documenting the applications themselves — screens, pages, endpoints, jobs go to application-documentation.
  • Not to pad output with narration or restate canonical rules.

Inputs

  • The fact/decision/state to document and its canonical home.
  • Existing docs that may already cover it.

Discovery Questions

  • Is this canonical (../../system/), project state (../../projects/current/), knowledge (../../knowledge/), or generated (../../generated/)?
  • Or is it a fact about the built product? Then it is docs/, not .ai/ — hand to application-documentation.
  • Does a doc already cover this (update vs create)?
  • Who reads this next, and what do they need?

Responsibilities

  • Place docs in the correct canonical location (see the map in DOCUMENTATION_RULES.md).
  • Keep adapters thin — link to .ai/, never duplicate it.
  • Record decisions and their reasons (including rejected alternatives).
  • Keep progress/task notes concise (status, decisions, links).
  • Update relevant indexes (e.g. ../README.md) when adding entries.
  • Archive stale generated content (../../generated/).

Required Workflow

  1. Identify the canonical home.
  2. Check for an existing doc to update.
  3. Write concisely; capture the "why."
  4. Update the relevant index.
  5. Archive superseded generated docs.

Decision Rules

  • Single source of truth: document once, link elsewhere.
  • Update over duplicate; delete/correct stale over contradict.
  • Generated reports are archivable, not permanent context.

Rules

  • No duplication of system rules into adapters or other docs.
  • No secrets/unredacted PII in any doc (../../system/SECURITY_RULES.md).
  • Write for reload: self-locating, with links.

Anti-Patterns

  • Fat adapters restating the system.
  • Verbose progress logs that bury the decision.
  • Duplicate docs that drift out of sync.
  • Long code examples in permanent skills (use ../../references/ / ../../templates/).

Validation Checklist

  • [ ] Correct canonical location.
  • [ ] Concise; decisions + reasons recorded.
  • [ ] No duplication of canonical content.
  • [ ] Index updated.
  • [ ] Stale generated docs archived.
  • [ ] No secrets/PII.

Definition of Done

Documentation placed canonically, concise, non-duplicative, with decisions and reasons recorded, indexes updated, and stale generated content archived.

Related Skills

application-documentation (the docs/ half), project-orchestrator, code-review, final-quality-audit, and any skill producing records.

Related Knowledge

../../knowledge/ (where durable knowledge lands).

Related References

../../references/ structure conventions.

Context Loading Guidance

  • Requires: the item to document and its canonical home.
  • Does not require: unrelated source, the full reference tree, other skills' bodies.
  • May load: the target index/README to update.
  • Stop when: the doc + index are updated and stale content archived.

Token Efficiency Guidance

Write the minimum that serves the next reader. Link rather than paste. Summarize source into decisions.

Security

  • Security scanPassed, 90 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

7-day trend score
0

Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/ahtishamshahzad/agent_dev_flow/documentation/badge)](https://verified-skill.com/skills/ahtishamshahzad/agent_dev_flow/documentation)

Something wrong with this skill or its page? Tell us.