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 commands/iii-experimental/agentos/agentos-evolvegit clone --depth 1 https://github.com/iii-experimental/agentosWhat 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 | $0.00016 | $0.00091 |
| Opus 5 | $0.00008 | $0.00046 |
| Sonnet 5 | $0.00003 | $0.00018 |
| Haiku 4.5 | $0.00002 | $0.00009 |
Grade B, and why
agentos-evolve scanned grade B with 2 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Sends data to an external URLmediumData exfiltration
A POST to an outside endpoint may be telemetry or may be exfiltration; either way the mod talks to somewhere, and you should know where.
curl -s -X POST http://localhost:3111/api/evolve/generate -H 'Content-Type: application/json' -d '{"goal": "$ARGUMENTS", "name": "evolved-fn", "agentId": "claude-code"}' | jq . Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
curl -s -X POST http://localhost:3111/api/evolve/generate -H 'Content-Type: application/json' -d '{"goal": "$ARGUMENTS", "name": "evolved-fn", "agentId": "claude-code"}' | jq . What it actually says
Evolve a function:
curl -s -X POST http://localhost:3111/api/evolve/generate -H 'Content-Type: application/json' -d '{"goal": "$ARGUMENTS", "name": "evolved-fn", "agentId": "claude-code"}' | jq .
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.
- 2d ago First seen · 10 lines · 16 tokens per session scan B 5ff34a2671ac
agentos-evolve is a command published in the GitHub repository iii-experimental/agentos (162 stars, last pushed 3mo ago), licensed Apache-2.0. It adds 16 tokens to every session and 91 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it B with 2 findings (sends data to an external url, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other commands, from other repositories
perf-difficult-conversation
Three-conversations framework rehearsal (Stone/Patton/Heen — Harvard Negotiation Project) for high-stakes conversations — termination, demotion, hard feedback, conflict surfacing. Produces three-conversations breakdown (what-happened / feelings / identity), conversation script with multiple opening choices, SCARF…
culture-onboarding-90
Design a 90-day onboarding architecture for a role. First day (psychological safety + belonging activation), first week (relationship + role context + early wins), first month (skill ramp + integration + baseline), first quarter (autonomous performance + cultural integration check). Written-first elements for…
org-reorg-trauma-audit
Pre-reorg trauma audit and sequencing plan. Opens with the 70%+ reorg-failure honesty, runs trauma history + threat-activation forecast (SCARF), produces a wave-based sequencing plan, communication architecture grounded in Kotter + Bridges, per-team mitigation, and post-reorg Talent IS monitoring cadence. Refuses…
design-regimen
Design an integrated weekly regimen — training, nutrition, sleep, stress, recovery — matched to the person's actual life. Composes with gym-training-expert and health-nutrition-expert for domain detail. Produces a sustainable architecture, not an extreme program. Not medical advice.
hire-debrief
Run the post-loop decision debrief — facilitator script for the 45-60 minute session. Surface inter-rater drift, anchor every claim to the rubric, name bias patterns out loud, render the hire-or-no-hire decision per the pre-committed rule, and capture learning for the next loop. Structured scores submitted BEFORE…
reclaim-knowledge
Organize scattered professional material into a domain-based second-brain architecture. Maps existing sources (Canva, Drive, local folders, Notion, screenshots) to functional domains. Non-technical-friendly — produces folder structure + drag-and-drop instructions, not CLI commands.