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/retrospective/agents-mdgit clone --depth 1 https://github.com/robcsaszar/retrospectiveWhat 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 | $0.00332 | $0.00332 |
| Opus 5 | $0.00166 | $0.00166 |
| Sonnet 5 | $0.00066 | $0.00066 |
| Haiku 4.5 | $0.00033 | $0.00033 |
Grade A, and why
retrospective 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 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.
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.
What it actually says
AGENTS.md
Mission
This repo publishes the retrospective skill for others to install. There is no build, no tests, no runtime of its own; the deliverable is the contents of skills/retrospective/. Judge changes by one test: would a stranger who installs this skill into their own project get a correct, generically useful session review out of it — one grounded in the conversation they are actually in?
Judgment boundaries
NEVER:
- Never let the skill read an on-disk transcript, log, or exported chat — its whole premise is that the current conversation is already the context. A version that opens a file to score it is a different, more dangerous skill.
- Never let the skill fabricate token counts, tool calls, or quotes. Relative magnitude only; the account is grounded in what happened, not in invented numbers.
ASK:
- Ask before adding a second skill; this repo is intentionally single-skill.
- Ask before restructuring the shape → context → wiring flow, or removing the honesty rules (current-session-only, no fabrication, never block on I/O).
- Ask before changing the license or copyright holder.
ALWAYS:
- Keep the skill's content generic — it must read as session review guidance for any project on any Agent Skills host, never tied to one codebase, tracker, or tool's command names.
- Keep the six shape names, the five economy-rule names, and the score key consistent between
SKILL.mdand thereferences/files — the review cites them by name, so a rename in one place without the other breaks the citations.
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 · 21 lines · 332 tokens per session scan A 6418d98e1de4
retrospective AGENTS.md is an instructions file published in the GitHub repository robcsaszar/retrospective (0 stars, last pushed 3d ago), licensed MIT. It adds 332 tokens to every session, about $0.0017 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
reflexio AGENTS.md
Instructions for ReflexioAI/reflexio, covering reflexio and quick reference.
BEST-Self-Enhancement-Learning-AI CLAUDE.md
Instructions for subkoks/BEST-Self-Enhancement-Learning-AI, covering bsela — claude code entry, completed, repo state at session end, next session — start here and state at session start.
AutoTrainess AGENTS.md
Instructions for simple-agent-lab/AutoTrainess, covering agents.md, objective, hard constraints, experiment guidance and stage rules.
smarter-agent AGENTS.md
Instructions for AnimaApp/smarter-agent: When you make a meaningful behavioral change to an agent skill, durable memory convention, system prompt, routing policy, safety boundary, classification rule, evidence threshold, escalation rule, or completion rule, load and follow skills/smarter-agent/SKILL.md before claiming…
BEST-Self-Enhancement-Learning-AI AGENTS.md
Instructions for subkoks/BEST-Self-Enhancement-Learning-AI, covering bsela — project agent rules, project identity, core invariants, memory taxonomy (canonical) and improvement loop contract.
socratic AGENTS.md
AGENTS.md instructions for robcsaszar/socratic, covering agents.md, mission and judgment boundaries.