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 skills add avmnu-sng/sutra --skill retro-and-memorygit clone --depth 1 https://github.com/avmnu-sng/sutraWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/avmnu-sng/sutra/retro-and-memory)<a href="https://agentmods.dev/skills/avmnu-sng/sutra/retro-and-memory"><img src="https://agentmods.dev/badge/skills/avmnu-sng/sutra/retro-and-memory/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/avmnu-sng/sutra/retro-and-memory"><img src="https://agentmods.dev/badge/skills/avmnu-sng/sutra/retro-and-memory.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What 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 5.1 | $0.00034 | $0.01623 |
| Opus 5 | $0.00017 | $0.00812 |
| Sonnet 5 | $0.00007 | $0.00325 |
| Haiku 4.5 | $0.00003 | $0.00162 |
Grade A, and why
retro-and-memory scanned grade A with 1 finding 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 10d 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.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
- Good: "New endpoint works -- `curl -s localhost:8080/health` -> `{"ok":true}`; How it starts
The opening of the file, as written. The whole thing — 155 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Retro and memory
A session ends. The instinct is to write "everything is green, shipped X, tests pass" and move on. That sentence is worthless the moment it is challenged, because none of it is backed by something a future session can re-run. The retro loop replaces claims with probes, reconciles outside critique against what the code actually does, and writes only the durable residue into whatever notes you keep, kept deliberately small.
The core is the per-session retro: before you call a session done, probe every health claim, audit what you actually shipped, and reconcile outside critique. What survives -- and only that -- is worth writing down. Where and how you persist it is your choice; the appendix sketches one optional shape.
When to use
- Closing out a work session that produced code, decisions, or findings.
- Recording a lesson you want the next session to inherit.
- Absorbing an outside review, adversarial critique, or reviewer's list of concerns.
- Optionally, on a periodic pass that rolls recurring findings into standing themes.
When not to
- Mid-session scratch notes. Retro is a close-out ritual, not a running log.
- A one-off fact with no reuse value. Not every observation earns an entry.
The per-session retro
1. Back every health claim with a runnable probe
No bare status claims. Each line that asserts the state of the world carries the exact command that proves it and the observed result -- so the next session re-runs the probe instead of trusting your prose.
-
Bad: "Test suite passes."
-
Good: "Test suite passes --
task test:unit-> 214 examples, 0 failures (run at close)." -
Bad: "The new endpoint works."
-
Good: "New endpoint works --
curl -s localhost:8080/health->{"ok":true}; drove the full create->read flow, not just the unit test."
Rules:
- A probe is a command anyone can paste, plus its observed output. "I checked" is not a probe.
- Prefer end-to-end drives over test-only evidence. A green test proves the test is green; it does not prove the feature works. Where it matters, exercise the real flow.
- If a claim has no probe, either find one or downgrade the claim to an open question. Do not ship an unbacked green.
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.
- 10d ago First seen · 155 lines · 34 tokens per session scan A 3f1f13a3654d
retro-and-memory is a skill published in the GitHub repository avmnu-sng/sutra (2 stars, last pushed 1mo ago), licensed MIT. It adds 34 tokens to every session and 1,623 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
report
Writes the session final report to a file, then prints only the path and a one-line summary. Fires when the prompt contains "Report per memstack:report", and also when the prompt begins with a standing trigger configured through MEMSTACKREPORTONTASKPROMPTS or MEMSTACKREPORTTRIGGERS. Dormant otherwise.
token-optimization
Use when the user says 'token optimization', 'save tokens', 'context window', 'reduce tokens', 'token stack', or 'TokenStack', or asks about extending context window capacity. Covers TokenStack, the built-in compression proxy that shrinks Claude Code tool output before it reaches the Anthropic API. Do NOT use for…
compress
Use when the user says 'tokenstack', 'compression', 'token savings', 'proxy status', or asks about context window usage.
state
Use when the user says 'update state', 'project state', 'where was I', or at session start to load current context.
grimoire
Use when the user says 'update context', 'update claude', 'save library', or after significant project changes.
catchup
Restore context after /clear by summarizing recent work and project state.