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 TAKEOFF69/codex-skills-kit --skill retro-distillgit clone --depth 1 https://github.com/TAKEOFF69/codex-skills-kitWrote 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/takeoff69/codex-skills-kit/retro-distill)<a href="https://agentmods.dev/skills/takeoff69/codex-skills-kit/retro-distill"><img src="https://agentmods.dev/badge/skills/takeoff69/codex-skills-kit/retro-distill/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/takeoff69/codex-skills-kit/retro-distill"><img src="https://agentmods.dev/badge/skills/takeoff69/codex-skills-kit/retro-distill.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 | $0.00104 | $0.03917 |
| Opus 5 | $0.00052 | $0.01959 |
| Sonnet 5 | $0.00021 | $0.00783 |
| Haiku 4.5 | $0.00010 | $0.00392 |
Grade A, and why
retro-distill 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 8d 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 — 294 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Self-Distillation Retrospective
You run structured retrospectives on recent AI coding-agent sessions to extract reusable patterns, anti-patterns, and workflow improvements. This is your self-improvement loop – inspired by simple self-distillation (SSD), where a model improves by sampling its own outputs, truncating the bad tails, and reshaping what remains.
The loop: Sessions produce outputs (samples) --> retro identifies what worked and what didn't (truncation) --> learnings get codified into skills, memories, and rules (reshaping) --> future sessions benefit.
Before Starting
1. Determine retro mode
| Mode | Trigger | Depth | Time |
|---|---|---|---|
| Quick | User says "retro", or a major chunk or failure merits learning capture | Review current session only | 2-3 min |
| Deep | After 5-10 sessions, user says "deep retro" or "distill" | Batch review of recent session logs | 10-15 min |
| Failure | Something went wrong, user says "failure retro" or "what went wrong" | Root-cause analysis of a specific failure | 5-10 min |
Default to Quick unless the user specifies otherwise.
2. Gather inputs
Quick mode:
- Review the current conversation context
- Check recent git log for what was committed
Deep mode:
- Read the last 5-10 session logs from your session-logs directory (e.g.
{your session-logs dir}) - Search relevant feedback memories by touched domain and failure pattern; do not bulk-load unrelated history
- Read the current anti-pattern lists in any skill
references/
Failure mode:
- Get the specific failure context (error, bad output, wasted effort)
- Read the session log where it happened
- Read the relevant skill/prompt that produced the failure
Quick Retro Protocol
Run through these four questions for the current session:
Q1: What worked? (Fork preservation)
Identify approaches that produced good results – especially non-obvious ones the system doesn't already codify. These are "forks" that led to correct solutions.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 8d ago First seen · 294 lines · 104 tokens per session scan A 60ab43e22fd7
retro-distill is a skill published in the GitHub repository TAKEOFF69/codex-skills-kit (7 stars, last pushed 1mo ago), licensed MIT. It adds 104 tokens to every session and 3,917 once invoked, about $0.0005 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 skills, from other repositories
agents-generator
Generate project-specific AGENTS.md and companion rules by analyzing a codebase. Supports full, minimal, update, and dry-run modes with package-manager detection, monorepos, backups, managed blocks, confidence scoring, and command validation.
nopus-configure
Configure nopus complexity sensitivity, extra-simple rewrites, rewrite evidence, and Pi response hiding for this user.
rescue
Delegate a substantial diagnosis, implementation, or follow-up task to Claude Code through the tracked-job runtime. Args: --background, --wait, --resume, --resume-last, --fresh, --write, --model , --effort , --prompt-file , [task text]. Defaults to opus with no forced effort. Use when Claude should investigate or…
adversarial-review
Run a design-challenging Claude Code review of local git changes in this repository. Args: --wait, --background, --base , --scope , --model , --effort , [focus text]. Defaults to opus with no forced effort. Use only when the user wants stronger scrutiny than a normal review, such as explicit tradeoff challenge…
review
Run a standard Claude Code review of local git changes in this repository. Args: --wait, --background, --base , --scope , --model , --effort . Defaults to opus with no forced effort. Use as the default path for ordinary code-review requests when the user did not explicitly ask for stronger adversarial scrutiny or for…
setup
Check whether Claude Code CLI is ready in this environment and optionally toggle the turn-end review gate. Args: --enable-review-gate, --disable-review-gate. Use for installation, authentication, or review-gate setup requests.