Skills

Generated from the repo's README by scripts/sync-wiki.sh. Edit the README, not this page.

Skills

Grouped into buckets under skills/; each bucket has a README.md listing its skills. Invoke as /context-forge:<skill> (Claude Code), $<skill> (Codex), or just describe what you want. Skills marked CC act on Claude Code's settings.json, hooks or status line and are Claude Code only; everything else is portable. Skills marked user-invoked write your config and never auto-trigger.

context — shrink what the agent loads or emits

Skill	What it does
optimize-claudemd	Rewrites a bloated CLAUDE.md under 300 words without dropping constraints
estimate-tokens	Per-source token estimate with instruction-file size thresholds
low-token-mode	Terse response discipline for the rest of the session
reset-context	Safely resets context window when near full — prevents hard stops
manage-skills CC	Audits loaded skills, disables unused ones — reduces session overhead
token-statusline CC, user-invoked	Adds live context bar to Claude Code status line

config — settings.json and hooks, documented keys only (CC, all three)

Skill	What it does
tune-settings user-invoked	Diffs, backs up and applies token-saving settings (autoMemoryEnabled, disableBundledSkills, disabledMcpjsonServers, autoCompactWindow)
project-isolation user-invoked	Scopes skills and hooks to current project only
debug-hooks	Diagnoses broken hook configurations with detailed validation output

productivity — workflow

Skill	What it does
task-brain-lite	Decomposes complex tasks into prioritized, dependency-aware subtasks
llm-wiki	Builds a persistent wiki the agent reads instead of re-ingesting raw docs each session

Agent: hook-error-fixer — diagnoses and auto-fixes broken hook configurations (backs up every file it edits).

---