retro

A short review process for examining unwanted behaviour by a coding agent and proposing a change to prevent it happening again.

In plain words
What is it for?
Use it after an agent skips discovery, starts coding too early, ignores project knowledge, or fails to pause; record approved fixes in project instructions or knowledge.
Why use it?
It turns a mistake into a documented correction instead of leaving the same instruction gap or workflow problem unresolved.

Command for Claude Code

Install

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.

agentmods
npx agentmods add commands/graph-tl/graph/retro
Clone the repo
git clone --depth 1 https://github.com/Graph-tl/graph

Made for: Claude Code.

Per session 0 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 224 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00000 $0.00224
Opus 5 $0.00000 $0.00112
Sonnet 5 $0.00000 $0.00045
Haiku 4.5 $0.00000 $0.00022

Measured yesterday against content hash 8521a018d707, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

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 yesterday.

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.

.claude/commands/retro.md · 17 lines

What it actually says

The user has identified unwanted agent behavior in this session. Run an ad-hoc retro to capture it.

Steps

  1. Identify the behavior — Review what just happened in this conversation. What did the agent do wrong? Common issues: skipping discovery, jumping to implementation without adding to graph, not pausing after tasks, ignoring graph workflow, not checking knowledge first.

  2. Root cause — Why did it happen? Missing instruction in CLAUDE.md? Ambiguous rule? Agent prompt gap? Convention not enforced by tooling?

  3. Propose a fix — Draft the exact CLAUDE.md instruction or graph knowledge entry that would prevent this in future sessions. Present it to the user for approval.

  4. Record it — Based on user approval:

    • Add to CLAUDE.md if it's a behavioral rule the default agent should always follow
    • Write as graph knowledge if it's project-specific context
    • Add as a graph node via graph_plan if it needs tooling changes
  5. Confirm — Show the user what was recorded and where.

Changes

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.

  1. yesterday First seen · 17 lines · 0 tokens per session scan A 8521a018d707

Subscribe to this mod's changes

retro is a command published in the GitHub repository Graph-tl/graph (6 stars, last pushed 6mo ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 224 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 2026-08-31.