Harness Forge
Description
One /forge command installs a self-improving Claude Code harness in any project: self-healing failure ledger, per-file lesson memory, regression evals, small-model elevation agents, and deterministic safety hooks.
Installation
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open the source below and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
README
harness-forge
One slash command that installs — and later upgrades — a **self-improving Claude Code harness** in any project. The harness it forges makes agents (including small/fast models like Haiku) safer, cheaper, and measurably more reliable, and it gets better over time by converting its own failures into durable, promoted rules.
Want to contribute? See [ROADMAP.md](ROADMAP.md) for independent, self-contained feature write-ups — pick one and you'll know exactly what's being asked and why.
What the forged harness contains
| Layer | Mechanism |
|---|---|
| Self-healing | Every tool failure is auto-recorded with a normalized signature (capture-failure.sh → machine-local ledger); repeats surface at session start with a prompt to convert them into lessons. |
| Continuous learning | Per-file lesson store with trigger conditions, verbatim evidence, and a vote lifecycle. /learn classifies each signal (explicit correction / implicit reframe / one-off mood / self-induced / positive confirmation — each its own weight) before ADD/UPDATE/UPVOTE/DOWNVOTE, promotes at weight ≥3.0 into hooks or regression checks, and applies an asymmetric trust gate when consuming a lesson: an added suggestion needs a citation, a blocked one needs an anchor (a rule conflict, a weight ≥3.0 lesson, or a deterministic check) — a soft match is surfaced to you, never auto-applied. Top-3 lessons injected each session. |
| Self-maintenance | /harness-audit reads a living per-project references/baseline.md first (source-of-truth table, wiring seams, known gotchas, a change log it appends to every run), then fixes mechanical invariants in-pass — including an issues-solved numbering-collision lint, a plans/ archival check, and a generic state-file seam-check (every state/ writer needs a reader). A regression-eval folder turns every confirmed failure into a permanent runnable check; loop.md routes bare /loop through finish-work → learn → audit → next task. |
| **Au |
Related Skills
Agency Agents
A complete AI agency at your fingertips - From frontend wizards to Reddit community ninjas, from whimsy inject
Awesome Llm Apps
100+ AI Agents, Agent Skills and RAG Apps - Free and Open Source.
Firecrawl
🔥 The API to search, scrape, and interact with the web for AI
Artifacts Builder
Suite of tools for creating elaborate, multi-component claude.ai HTML artifacts using modern frontend web tech
Headroom
Compress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agen
CrewAI
Framework for orchestrating role-playing, autonomous AI agents. By fostering collaborative intelligence, CrewA
