ccc-compound

ccc-compound is a command for coding agents from KevinZai/commander. It costs 19 tokens per session (862 once invoked), scanned A, original, MIT.

Post-task learning capture — extract patterns, corrections, and decisions to compound productivity.

Command

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/kevinzai/commander/ccc-compound
Clone the repo
git clone --depth 1 https://github.com/KevinZai/commander

Wrote 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.

agentmods badge for ccc-compound

README.md
[![agentmods](https://agentmods.dev/badge/commands/kevinzai/commander/ccc-compound.svg)](https://agentmods.dev/commands/kevinzai/commander/ccc-compound)
Your own site
<a href="https://agentmods.dev/commands/kevinzai/commander/ccc-compound"><img src="https://agentmods.dev/badge/commands/kevinzai/commander/ccc-compound.svg" alt="Measured on agentmods" height="20"></a>
Per session 19 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 862 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin unknown 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 $0.00019 $0.00862
Opus 5 $0.00010 $0.00431
Sonnet 5 $0.00004 $0.00172
Haiku 4.5 $0.00002 $0.00086

Measured today against content hash 2463ed426d82, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

ccc-compound 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 today.

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.

commands/ccc-compound.md · 124 lines

How it starts

The opening of the file, as written. The whole thing — 124 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Compound — Post-Task Learning Capture

Inspired by Compound Engineering by Every.to

Each unit of work makes the next one slightly cheaper. Extract the signal, persist it, move on.

Argument: {{input}}

  • No argument = capture learnings from the current session
  • review = display all accumulated lessons without adding new ones
  • prune = review lessons and remove outdated or irrelevant entries

If review

Read tasks/lessons.md and ~/.claude/learned-skills/ contents. Display a summary:

COMPOUND REVIEW

Lessons: X entries across Y categories
Learned Skills: Z entries

[grouped list of lessons by date, most recent first]

Stop here. Do not extract or persist anything.

If prune

Read tasks/lessons.md. For each entry, evaluate whether it is still relevant given the current codebase and project CLAUDE.md. Present entries to remove with rationale, then ask for confirmation before deleting.

Stop here after pruning.


Default: Capture Learnings

1. Review What Just Happened

Run in parallel:

  • git log --oneline -10 — recent commits
  • git diff --stat HEAD~3 — recent file changes

Then review the conversation context for decisions made, corrections received, and patterns discovered.

2. Extract Learnings

Organize into three categories:

Patterns Discovered:

  • Code patterns that worked well (architecture, data flow, API design)
  • Effective tool or library usage worth repeating
  • Workflow sequences that were efficient

Mistakes & Corrections:

  • What went wrong and the root cause
  • What the fix was
  • How to prevent this class of error in the future

Decisions & Rationale:

  • Key decisions made and WHY (not just what)
  • Alternatives considered and rejected
  • Constraints that drove the decision

Skip any category with nothing meaningful to report.

3. Persist

  • Append actionable entries to tasks/lessons.md (create the file if it does not exist). Use the format:
    ## YYYY-MM-DD — [session summary]
    
    ### Patterns
    - [pattern]
    
    ### Corrections
    - [mistake] -> [fix] -> [prevention rule]
    
    ### Decisions
    - [decision]: [rationale]
    
  • If a pattern is broadly reusable, suggest adding it to the project CLAUDE.md.
  • If a correction reveals a gap in existing rules, suggest a specific new rule.
  • Check ~/.claude/learned-skills/ and suggest new learned skill entries if appropriate.

Read the full file on GitHub · 124 lines

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. today First seen · 124 lines · 19 tokens per session scan A 2463ed426d82

Subscribe to this mod's changes

ccc-compound is a command published in the GitHub repository KevinZai/commander (6 stars, last pushed yesterday), licensed MIT. It adds 19 tokens to every session and 862 once invoked, about $0.0001 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 2026-09-03.