Install
$ agentstack add skill-fazalrshah-claude-skills-living-docs-workflow ✓ scanned · ✓ verified — works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
About
Living-Docs Workflow
A discipline for AI-assisted projects that outlive a single context window: nothing lives in the model's memory — everything lives in versioned markdown. Any fresh chat reads a fixed set of files and is instantly caught up.
The files (create these at the project root)
ARCHITECTURE.md— the single source of truth: the system map, locked decisions, stack, and gotchas.
If a decision changes, update THIS file (not just chat) so future chats inherit it.
CLAUDE.md— the pickup rules: an ordered "read these first in every new chat" list pointing at the
files below, plus auto-save rules. This is what makes resume reliable.
session-logs/session.md— append-only running log: a timestamped entry after every meaningful action
(code written, command run, bug fixed, decision made). Never overwrite.
issues.md— observed bugs asISS-NNN, each OPEN until re-verified (🔴/🟡/🟢 with date + evidence).notes.md— improvement backlog / ideas, tagged to the relevant component, surfaced on demand.tests.md— acceptance/verification checklist to run when an area is "done".- (optional)
Contracts.md,security.md,rag.md,opensource-ideas.mdfor domain-specific tracks.
Pickup rules (put in CLAUDE.md)
- Read
ARCHITECTURE.md(source of truth). 2. Read the tail ofsession-logs/session.md. 3. ReadCLAUDE.md. - Read
notes.md(surface relevant items per component). 5. Readissues.md(re-verify open issues in-area).
Add a one-line PENDING PRIORITY banner at the top of CLAUDE.md when a specific next task is agreed, so the next chat surfaces it even if the user opens a different topic.
Session snapshots (resume)
On "save session", write a dated snapshot to a snapshots dir capturing: what worked (with evidence), what failed/gotchas, exact next step, and environment notes. On "resume", load the latest snapshot.
Rules that make it work
- Update the doc, not just chat. A decision only spoken in chat is lost next session.
- Append, never overwrite the session log.
- Issues stay OPEN until re-tested — flip to 🟢 only with date + evidence.
- Capture, don't carry. The moment you notice a reusable lesson or a bug, write it to the right file.
Why it works
Multi-chat AI builds fail when context is lost between windows. A fixed "read these N files first" contract + append-only logs + dated snapshots means every instance is fully current in seconds, decisions never drift, and bugs don't silently reappear.
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: fazalrshah
- Source: fazalrshah/claude-skills
- License: MIT
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet — be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.