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/chankov/agent-fleet/compound-learningnpx skills add chankov/agent-fleet --skill compound-learninggit clone --depth 1 https://github.com/chankov/agent-fleetWhat 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 | $0.00091 | $0.02265 |
| Opus 5 | $0.00046 | $0.01132 |
| Sonnet 5 | $0.00018 | $0.00453 |
| Haiku 4.5 | $0.00009 | $0.00227 |
Grade A, and why
compound-learning 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 2d 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 — 138 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Compound Learning
Overview
Every session produces knowledge that is more durable than the diff: corrections the user had to make twice, review findings that keep recurring, root causes that took an hour to find, assumptions that turned out wrong. When the session ends, that knowledge evaporates — and the next session pays for it again.
This skill is the compound step: it converts session evidence into small, deduplicated updates to the project's own rule tree (HOW — compliance) and documentation (WHAT/WHY — orientation), so each session makes the next one better.
The enemy is not missing lessons — it is bloat. A rule tree that grows by five generic rules per session becomes noise that future sessions skim past. Every part of this process is biased toward fewer, sharper updates on existing files: the default verdict for any candidate lesson is "not worth keeping".
When to Use
- A session is wrapping up and it contained at least one correction, recurring finding, or hard-won discovery.
- The user says "compound", "capture the lessons", "update the rules from this session", or runs a
/compoundcommand. - You are the
documenterpersona dispatched with a compound task by the agent-hub dispatcher. - Periodically, as a consolidation pass over a rule tree that has accumulated additions (see Step 8).
When NOT to use: mid-task (finish the work first — compounding interrupts flow and the evidence isn't complete); after a trivial session with nothing corrected or discovered; to document what the code or git history already records.
Where Evidence and Targets Live
Evidence — read what exists, skip what doesn't:
| Source | Where |
|---|---|
| Session conversation | User corrections, "no, do it this way", decisions, rejected approaches |
| Session diff | git diff / git log for the session's commits |
| agent-hub artifacts | .pi/agent-sessions/artifacts/{returns,reviews,plans,inventories,evidence}/ — specialist returns and review findings |
| agent-hub assertion ledger | .pi/agent-sessions/assertions.json — which assertions failed first and why |
| Dispatch brief | A /compound dispatch carries a candidate-lessons brief composed by the dispatcher — treat it as candidates, not conclusions |
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.
- 2d ago First seen · 138 lines · 91 tokens per session scan A 4e6490e81fe1
compound-learning is a skill published in the GitHub repository chankov/agent-fleet (10 stars, last pushed 7d ago), licensed MIT. It adds 91 tokens to every session and 2,265 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
systematic-debugging
Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes.
brainstorming
You MUST use this before any creative work - creating features, building components, adding functionality, or modifying behavior. Explores user intent, requirements and design before implementation.
auto-perf-optimize
Run agent-driven VS Code performance or memory investigations. Use when asked to launch Code OSS, automate a VS Code scenario, run the Chat memory smoke runner, capture renderer heap snapshots, take workflow screenshots, compare run summaries, or drive a repeatable scenario before heap-snapshot analysis.
chat-perf
Run chat perf benchmarks and memory leak checks against the local dev build or any published VS Code version. Use when investigating chat rendering regressions, validating perf-sensitive changes to chat UI, or checking for memory leaks in the chat response pipeline.
chat-pet-sprite-creation
Use when creating or changing VS Code chat pet sprite art, sprite sheets, state animations, eye treatments, Stable/Insiders variants, or pet transitions under src/vs/workbench/contrib/chat/browser/widget/media/chatPet.
cpu-profile-analysis
Analyze V8/Chrome CPU profiles (.cpuprofile) and DevTools trace files (Trace-.json). Use when: profiling performance, investigating slow functions, comparing code paths, finding bottlenecks, analyzing timeToRequest, understanding call trees from sampling profiler data, analyzing layout/paint/rendering, investigating…