Skip to content
Back to skills

Context Management

ASecurity

Maintain and compress the context store so analysis is done once and reused, never repeated. Use to record a confirmed root cause, an architecture decision, or a service/contract change; or to compress sprawling notes back into the terse store. Triggers on "remember this", "update the context", "record decision", "we already analyzed this", "compress".

  • 3 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added September 24, 2026
ai-agentsrustgogit

Security analysis

A100/100

Scanned September 24, 2026

npx -y skills add OneDro1d/dark-factory --skill context-management --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Context Management?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Context Management
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/onedro1d-context-management/badge)](https://www.skillsdirectory.com/skills/onedro1d-context-management)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: context-management
description: Maintain and compress the context store so analysis is done once and reused, never repeated. Use to record a confirmed root cause, an architecture decision, or a service/contract change; or to compress sprawling notes back into the terse store. Triggers on "remember this", "update the context", "record decision", "we already analyzed this", "compress".
allowed-tools: Read, Write, Edit, Grep, Glob
---

# Context Management (analyze once, reuse forever)

Goal: keep `.claude/context/` the single, terse, trustworthy memory so agents never re-scan the repo to re-learn what's known. This is the main token-saving lever.

## The store
- `SERVICE-MAP.md` — components + flows (one line each).
- `FINDINGS.md` — confirmed bugs/root causes (append-only ledger, status-tagged).
- `DECISIONS.md` — architecture decisions/rulings (ADR log).
- `<service>/CLAUDE.md` — deep per-service detail.
- shared-models `CLAUDE.md` — contract index.

## When to write
- **Confirmed root cause** → new `F-###` in FINDINGS.md (symptom, cause, `file:line`, status). So it's never re-investigated.
- **Design/architecture decision** → new `D-###` in DECISIONS.md (decision, why, source).
- **New/changed service, routing key, table, object-store key** → update SERVICE-MAP.md + the service CLAUDE.md + (if a contract) the shared-models index.
- **Status change** (OPEN→FIXED) → update the existing entry, don't duplicate.

## How to write (compression rules)
- One fact per entry. Cite a source (`file:line`, ruling, ticket) for every claim — no guessing.
- Prefer tables/IDs over prose. Replace, don't append, when a fact changes.
- If notes have grown long, distill to the durable fact and delete the scratch.
- Keep each context file roughly a screen — if it's growing, compress.

## Read-first reminder for every agent
SERVICE-MAP → FINDINGS → DECISIONS → the one relevant CLAUDE.md → source. Stop as soon as you have enough.

## Don't
- Don't record session chatter or anything the code/git already states.
- Don't let the store drift from code — if you find a stale entry, fix it (cite the code).

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…