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 skills/donald-ada/workinggenius/distillnpx skills add donald-ada/workinggenius --skill distillgit clone --depth 1 https://github.com/donald-ada/workinggeniusWrote 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/donald-ada/workinggenius/distill)<a href="https://agentmods.dev/skills/donald-ada/workinggenius/distill"><img src="https://agentmods.dev/badge/skills/donald-ada/workinggenius/distill.svg" alt="Measured on agentmods" 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 | $0.00037 | $0.00827 |
| Opus 5 | $0.00018 | $0.00413 |
| Sonnet 5 | $0.00007 | $0.00165 |
| Haiku 4.5 | $0.00004 | $0.00083 |
Grade A, and why
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 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.
What it actually says
Distill
Tenacity distills a log at close-out. Work that finished before that rule existed — or closed without it — still carries everything: per-criterion evidence for tests that live in the tree, full bodies of contracts the final version replaced, an interview whose confirmed problem has lived in the snapshot ever since. This command runs the same move, later, on the logs that never got their pass.
The concept: close-out's one rule — does the repo answer this now? — applied where close-out never was, with the check close-out didn't need. At close-out the verification is minutes old; here it may be months, so the rule is a live question per item, not an assumption:
- Scope: done, has a log, undistilled. Only works at
stage: donethat have a<slug>.log.mdbeside the snapshot whose first line is notdistilled. In-flight logs are never touched, whatever the argument says. A done work with no log at all has nothing to distill: skip it and say so. - An item leaves only if the repo still answers it. Per-criterion evidence goes where the test it names is still in the tree; a test the repo has since dropped makes that line possibly the last record standing — it stays. Superseded contract bodies go (each keeps its one-line why); the interview's play-by-play goes where the snapshot carries the confirmed problem; a close-out review report shrinks to its one-line stub where the snapshot carries its findings and resolutions. What code can never answer always stays: decisions and kill-reasons, corrections and what overturned them, the user's words as they said them.
- A pointed-at entry stays, whatever its category. Before an entry leaves, check nothing still links to it — ⚠
.genius/BACKLOG.log.mdfirst — a retired seed's body went in carrying itsFromlink, and unlike every other pointer here that one can never be repaired, because the file it sits in is append-only. Then.genius/DECIDED.md,.genius/BACKLOG.md, whose every line carries aFrom [<slug>](…#<anchor>)back into some log, and ⚠ the work's ownCONTRACT.md, whose established layer names its source entry per block. That last one is inside the folder and easy to read past, and it is the one that outlives the rest: the contract is never drained, not at a bump and not at done, so its[source]links are still live when this pass runs months later — and per-criterion evidence is exactly what a block established by a slice points at. An anchor something points at is load-bearing; deleting it breaks the door that pointer exists to be. - Propose, then delete — never the other way. Read-only until the user has seen, per log, what leaves and what stays (categories and counts, not a line-by-line ballot); one confirmation covers the batch. A deletion the user hasn't seen is a silent one, and this flow doesn't do silent removal (
errataskill). - Mark the pass. The log's first line becomes
distilled retroactively, <date>. Links that pointed at removed entries leave with them — a slice line keeps its check and its words, never a dead anchor — so/reconcilefinds no holes.
Done when every targeted log is distilled, or named as skipped with its reason. Run it once per backlog of old work; after that, close-out keeps every new log distilled on its way out.
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.
- yesterday Changed fabe7305a630
- 6d ago First seen · 21 lines · 37 tokens per session scan A 93303b42312a
distill is a skill published in the GitHub repository donald-ada/workinggenius (6 stars, last pushed yesterday), licensed MIT. It adds 37 tokens to every session and 827 once invoked, about $0.0002 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
dw-handoff
Compact the live session into .ai/work/ /HANDOFF.md — how far into the current task you are, what is applied but uncommitted, and which dead ends are already ruled out, so the next context window resumes mid-task. Explicit-invoke only.
memory-bank
Maintain and update project memory-bank files for persistent context across Claude Code sessions. Use when starting a new session, completing features, making architectural changes, or when context needs to be preserved. Triggers on: memory bank, project context, session start, context update, CLAUDE.md, project…
alive:session-history
Revive sessions (quick or heavy), browse, and search — 'what happened recently?', 'find the session where we discussed X', 'revive yesterday's session'. For single-session recall and multi-session browsing. If the human needs to merge multiple sessions into one working context or detect conflicts between parallel…
recall
Must be used at the start of any non-trivial task involving code changes, debugging, repo exploration, file inspection, or environment/tooling investigation to surface stored guidance before analysis or tool use.
cross-session-handoff
Read, write, snapshot, and lock .arcgentic/state.yaml across planner, dev, audit, and optional test sessions.
knowledge-priming-refiner
Facilitate a structured conversation to create a project-specific knowledge base document. Produces a knowledge-base.md that primes AI with the project's tech stack, architecture, trusted sources, and project structure. Use when the user says 'set up knowledge base', 'prime the project', 'onboard AI', 'create…