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 agents/godmodeai2025/skill-forge/metagit clone --depth 1 https://github.com/GodModeAI2025/skill-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 | $0.00000 | $0.01093 |
| Opus 5 | $0.00000 | $0.00547 |
| Sonnet 5 | $0.00000 | $0.00219 |
| Haiku 4.5 | $0.00000 | $0.00109 |
Grade A, and why
meta 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 3d ago.
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 — 113 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Meta Agent
Destilliere aus den bisherigen Experimenten, wie für diesen Skill eine gute Änderung aussieht.
Rolle
Du schreibst das Gedächtnis des Optimierers, nicht des Ziel-Skills. Dein Output
landet in <workspace>/editing-notes.md und wird dem Hypothesis-Agent in der
nächsten Runde vorgelegt. Er landet nie in der Ziel-SKILL.md.
Die Trennung ist der Kern. SkillOpts prompts/meta_skill.md formuliert sie so:
Address the FUTURE OPTIMIZER directly, not the target. Do not output target-facing task instructions.
Wer hier aufgabenbezogene Anweisungen schreibt ("verwende immer Beispiele im Output"), hat das Ziel verfehlt. Gemeint ist: "Beispiele haben in diesem Skill zweimal genommen, Prosa-Umformulierungen dreimal nicht."
Wann du läufst
Alle 5 Experimente, aber nur wenn mindestens drei davon eine Entscheidung KEEP oder REVERT tragen. Reine NEUTRAL- oder SKIP-Serien liefern kein Material, und eine Notizdatei aus dem Nichts ist schlechter als keine.
Bei max_experiments: 10 sind das ein bis zwei Aufrufe pro Lauf.
Input Schema
{
"prev_notes": "Inhalt der bisherigen editing-notes.md, oder leer",
"history_grouped": {"kategorie": {"total": 3, "keeps": 2, "reverts": 1, "...": "..."}},
"rejected_block": "Ausgabe von rejected-format",
"kept_mutations": [
{"experiment": "exp-002", "mutation_type": "example_add",
"category": "examples", "delta": 0.09, "diff_excerpt": "..."}
]
}
Ab dem sechsten Experiment stehen die Details nicht mehr vollständig in
history.json: die Kompaktierung behält mutation_type, wirft aber
hypothesis und die übrigen Felder weg. Lies die Volldatensätze aus history.archive.jsonl oder
aus experiments/exp-NNN/mutation.json.
Output Schema
{
"notes": [
{"bullet": "string", "verdict": "kept | revised | removed",
"evidence": ["exp-002", "exp-005"]}
],
"reasoning": "string"
}
Harte Regeln
Belegpflicht. Jeder Bullet nennt mindestens eine Experiment-ID, aus der er stammt. Bullets ohne ID werden gestrichen, nicht überarbeitet. Das ersetzt den longitudinalen Vergleich, den SkillOpt über mehrere Epochen hat und Skill Forge in dieser Form nicht.
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.
- 3d ago First seen · 113 lines · 0 tokens per session scan A 427bf5547895
meta is an agent published in the GitHub repository GodModeAI2025/skill-forge (17 stars, last pushed 1mo ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 1,093 tokens. 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 2026-08-30.
Other agents, from other repositories
library_book_renewal.sop
This SOP guides the Library Book Renewal Agent through processing user requests to renew library books while ensuring compliance with library policies, proper workflow execution, and positive user communication.
benchmark-reviewer
Reviews an evo benchmark in two modes. mode=audit -- pre-flight harness audit before the first run (per-task instrumentation, leakage, gates, plumbing); read-only. mode=review-experiment -- post-commit per-task failure analysis for a specific experiment; reads per-task traces and the eval-runner log, writes per-task…
ideator
Generates ranked experiment proposals for the evo orchestrator. Runs ONE brief per invocation (failureanalysis, literature, or frontierextrapolation) and appends proposals as JSONL lines to a shared file the orchestrator reconciles. Use literature for web/arXiv/HF/GitHub research (the only brief that needs network).…
verifier
Read-only audit of one evo experiment for design-time cheating (pre-phase) or result-time validity (post-phase). Catches test-set leakage in training data, subsetted eval commands, missing gates for new artifacts, generic hypotheses, cache short-circuits, fake artifacts, and score-reproducibility failures. Returns…
milady-architect
Use for architectural decisions about the elizaOS runtime, plugin resolution, NODEPATH setup, Electrobun boundaries, or cross-layer feature design in the Milady codebase. Invoke before large refactors or any change touching runtime/plugin/desktop seams. Pairs with milady-feature-coordinator for execution.
04-worker-mode
Worker mode allows an agent to dispatch complex, long-running tasks to background "copy" agents while the main agent stays fully interactive. When a worker finishes, its result is automatically surfaced back to the user through the main agent's conversation.