Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add instructions/robcsaszar/ai-forge/agents-mdgit clone --depth 1 https://github.com/robcsaszar/ai-forgeWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.01049 | $0.01049 |
| Opus 5 | $0.00524 | $0.00524 |
| Sonnet 5 | $0.00210 | $0.00210 |
| Haiku 4.5 | $0.00105 | $0.00105 |
Grade A, and why
ai-forge AGENTS.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 59 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
Mission
This repo publishes the ai-forge skill pack: eight skills covering the lifecycle of an AI skill or agent definition, create, judge, review, apply, evaluate, recap, update, and audit. There is no build, no tests, no runtime of its own. The deliverable is the contents of skills/*, each copied in from a source repo (Orakl) where they are used and refined, then published here for anyone to install. Changes should be judged by: would a stranger who installs one of these skills into an unrelated project get correct, generic guidance out of it?
Layout convention
- One directory per skill under
skills/<name>/,name:in frontmatter matching the directory name exactly. - Each skill's full directory ships as-is:
SKILL.mdplus anyreferences/,assets/,agents/,evals/, orscripts/it depends on. Don't split a skill across a partial copy. - Three skills ship a
scripts/directory:ai-forge-create/scripts/validate-metadata.cjs,ai-forge-audit/scripts/check-ecosystem.cjs, andai-forge-eval/scripts/aggregate-benchmark.cjs. All are zero-dependency Node scripts with no network calls. SeeSAFETY.mdfor the current, verified description of each. - Skill bodies stay under 200 lines and ~1,800 words. Reference files over 300 lines carry a
## ContentsTOC.
Judgment boundaries
NEVER:
- Never let a copied skill's content leak repo-specific references back to Orakl (or any other single source project). A published skill needs to read as generic guidance, not one project's internal notes.
- Never remove a skill without asking first: the eight here form one lifecycle, and removing one leaves a gap other skills' cross-references (e.g.
ai-forge-auditrequiringai-forge-judge) assume is filled.
ASK:
- Ask before changing the license or copyright holder.
- Ask before adding a ninth skill that doesn't fit the create/judge/review/apply/eval/recap/update/audit lifecycle the other eight cover.
ALWAYS:
- When a skill's
scripts/directory gains, loses, or changes what a script does: updateSAFETY.mdin the same change, based on actually reading the script, not assuming its behavior is unchanged. - When adding, removing, or renaming a skill: update the table in
README.mdin the same change. - Before publishing, run the three pre-publish checks below. All three must be clean.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- yesterday First seen · 59 lines · 1,049 tokens per session scan A 41fdad303949
ai-forge AGENTS.md is an instructions file published in the GitHub repository robcsaszar/ai-forge (0 stars, last pushed 3d ago), licensed MIT. It adds 1,049 tokens to every session, about $0.0052 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
apm python.instructions.md
Python development guidelines.
kglite CLAUDE.md
Instructions for kkollsga/kglite, covering kglite — claude code conventions, build & test, architecture, the boundary principle (wrappers vs core) — summary and in-memory is the core product.
openings-mcp AGENTS.md
Instructions for amikai/openings-mcp, a project described as: 💼 One MCP server to search job boards and company career sites.
Awesome-Prompt-Engineering contribution-review.instructions.md
Instructions for natnew/Awesome-Prompt-Engineering, covering reviewing pull requests and issues, tone, how to respond, explaining quality concerns and maintainer judgement.
paper-pilot CLAUDE.md
Instructions for aytzey/paper-pilot, covering claude.md, best fit, setup, first workflow to try and claude prompt patterns.
mcpelevator CLAUDE.md
Instructions for pacnpal/mcpelevator, covering commands, architecture, conventions, agent skills and issue tracker.