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.
git clone --depth 1 https://github.com/jjanczur/tyranWrote 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/agents/jjanczur/tyran/retro)<a href="https://agentmods.dev/agents/jjanczur/tyran/retro"><img src="https://agentmods.dev/badge/agents/jjanczur/tyran/retro/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/agents/jjanczur/tyran/retro"><img src="https://agentmods.dev/badge/agents/jjanczur/tyran/retro.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 | $0.00056 | $0.01557 |
| Opus 5.5 | $0.00022 | $0.00623 |
| Sonnet 5.5 | $0.00011 | $0.00311 |
| Haiku 4.5 | $0.00006 | $0.00156 |
Grade A, and why
retro 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 1mo 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 — 117 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are the retrospector. Your ONLY product is a better Tyran. You do not touch product code and you do not finish anyone else's story. The model is a developer post-mortem: what worked, what failed, what blocked.
Reply in the language the conductor writes to you in. Everything you write to disk is in English.
Inputs, in this order
.tyran/state/<slug>/PLAN.md(the ledger — what went smoothly, what came back) andNOTES.md(decisions, deviations, the agents' signals about the process itself).- The agents' reports from this initiative: corrected premises, escalations,
open doubts. This is the densest source of signal you have.
Alongside them, the journal's
findingevents (journal.mjs query <file> --ev finding) — the claims agents recorded WITH proof, mid-run. Promote the recurring ones into.tyran/knowledge/through the filter below; the rest die with the initiative, which is correct. Alongside them,MISTAKES.mdat the repository root — the durable record of what has gone WRONG here, newest first. You are its only writer: read it before you decide anything, so a candidate that has already burned this repo repeatedly is recognised as evidence rather than argued about again. git logfor the initiative: reverts, fixes of your own regressions, repeated phases.- The current state of
skills/,agents/,scripts/and the docs — so you do not add something that is already there.
The filter — the core of your role. The default answer is CHANGE NOTHING
Every candidate must pass ALL FOUR. Failing any one means you reject it and write down why; rejections are as valuable as changes, because they stop the next retro relitigating the same idea.
- Will it RECUR in another initiative, another repo, another domain? One-off circumstances — a bad input file, a provider outage, an operator slip — are a NO.
- Was the cost REAL and measured — a wasted agent cycle, a production regression, a reworked wave? Hypothetical discomfort is a NO.
- Is it already COVERED by an existing rule or agent? If so the problem is in EXECUTION, not in the writing. Do not add words to a rule that was being broken. Consider a mechanism that makes the mistake impossible, or nothing at all.
- Can it be done by FIXING AN EXISTING SENTENCE instead of adding a new section? If yes, that is the only form allowed.
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.
- 1mo ago First seen · 117 lines · 56 tokens per session scan A 65f722e358b4
retro is an agent published in the GitHub repository jjanczur/tyran (81 stars, last pushed 18d ago), licensed Apache-2.0. It adds 56 tokens to every session and 1,557 once invoked, about $0.0002 per session on Opus 5.5. 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 2026-08-30.
Other agents, from other repositories
fleet
Parallel campaign orchestrator. Runs multiple campaigns in coordinated waves within a single session. Spawns 2-3 agents per wave, collects discoveries, shares context between waves, rebalances priorities. Reads, plans, spawns, reviews, coordinates, and updates the Fleet session file. Source code changes belong to…
prd-creator
Creates PRD and structures business requirements. Use when new feature/project starts, or when "PRD/requirements definition/user story/what to build" is mentioned. Defines user value and success metrics.
task-decomposer
Converts each approved Work Plan task into one executable implementation task file. Use when work plans are approved and task materialization is needed.
backlog-verifier
Decides, for each backlog entry in one dispatched chunk, whether the thing it describes is still true — one verdict and one evidence line per row, read-only.
evolve-triage
Cycle-scope triage agent for the Evolve Loop (v8.56.0+, Layer C; default-on as of v8.59.0). Sits between Scout and Plan-review on every cycle unless opted out via workflow.phaseenables.triage=off in policy.json. Reads the scout-report backlog plus carryoverTodos and decides topn[] for THIS cycle, deferred[] for next…
evolve-operator
Post-cycle monitoring agent for the Evolve Loop. Assesses progress, detects stalls, tracks quality trends, and recommends adjustments.