Agent skill

Compound

by garagon in garagon/nanostack

Document what you learned during this sprint. An agent skill from garagon/nanostack.

Apache-2.0Auto-check passed

Install Compound

skills CLI
$ npx skills add garagon/nanostack --skill compound -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install garagon/nanostack compound --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/garagon/nanostack.git skills-src && mkdir -p .claude/skills && cp -r skills-src/compound .claude/skills/compound && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
compound
GitHub stars
207
Token cost
~2.2k tokens
SKILL.md length
931 words
Files
1
Skills in repo
14
Repo updated
First seen
Licence
Apache-2.0

At a glance

Document what you learned during this sprint. An agent skill from garagon/nanostack.

  • Works in 8 steps: Read the sprint artifacts → Identify what's worth capturing → Check for existing solutions → …
  • SKILL.md covers Telemetry preamble, When to run, Process and Telemetry finalize, plus 1 more section
  • Calls jq

What it does

Compound is an agent skill from garagon/nanostack. Document what you learned during this sprint. Reads artifacts, writes structured solutions to know-how/solutions/. Run after /ship or after fixing a significant bug. Triggers on /compound.

Its SKILL.md is about 2.2k tokens, which your agent loads only when the skill is triggered. It is a single SKILL.md file with no bundled scripts.

The repository describes itself as: A workflow harness that helps AI coding agents plan, review, test, and ship safer code. The licence is Apache-2.0.

Example prompts

  • “/compound”

Workflow steps

8 steps, taken from the step headings in SKILL.md.

  1. Read the sprint artifacts
  2. Identify what's worth capturing
  3. Check for existing solutions
  4. Write solution documents
  5. Update frontmatter
  6. Report
  7. Check graduation candidates
  8. Check stale diarizations

What it can do on your machine

Read from SKILL.md and the folder at commit 0372aed. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    Shell commands in SKILL.md call:

    • jq

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Compound loads about 2.2k tokens when it runs. Until then it costs about 49 tokens; SKILL.md has 931 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~49
When it runs · the whole SKILL.md, loaded when a task matches
~2.2k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from garagon/nanostack at commit 0372aed, republished under its Apache-2.0 licence (© garagon). 931 words, ~2,247 tokens.

Download SKILL.mdSave it as .claude/skills/compound/SKILL.md (or your agent's skills folder).
name
compound
description
Document what you learned during this sprint. Reads artifacts, writes structured solutions to know-how/solutions/. Run after /ship or after fixing a significant bug. Triggers on /compound.
concurrency
write
depends_on
ship
summary
Knowledge capture. Documents bugs, patterns, decisions from sprint artifacts.
estimated_tokens
250

/compound - Knowledge Compounding

After a sprint or a significant fix, extract what you learned into structured, searchable documents. Next time the agent plans or reviews, it finds these automatically.

Telemetry preamble

Defensive telemetry init. No-op if telemetry is disabled via NANOSTACK_NO_TELEMETRY=1, ~/.nanostack/.telemetry-disabled, or if the helpers are removed.

bash
_P="$HOME/.claude/skills/nanostack/bin/lib/skill-preamble.sh"
[ -f "$_P" ] && . "$_P" compound
unset _P

When to run

  • After /ship completes a sprint
  • After fixing a bug that took significant investigation
  • After making an architecture decision worth remembering
  • After discovering a pattern that should be reused

Process

1. Read the sprint artifacts

Resolve all sprint phase artifacts in one call:

bash
~/.claude/skills/nanostack/bin/resolve.sh compound

The output is JSON with upstream_artifacts containing paths for think, plan, review, security, qa, and ship artifacts (only those that exist within the last 2 days). Read what's available. Focus on:

  • /review findings that were fixed (these are bugs worth documenting)
  • /security findings that were resolved (these are patterns worth remembering)
  • /think scope decisions (these are decisions worth recording)
  • /qa failures that were debugged (these are bugs with investigation trails)

Report-only ship check. Read the ship artifact's run_mode. If it is "report_only", /ship only previewed the delivery and nothing was published (no commit, no PR, no deploy). Tell the user plainly that the ship was report-only, then keep going: the investigation from /review, /security, and /qa is still worth capturing if it was non-trivial. But do not write a solution or summary that claims work was shipped or that the sprint closed. There is no shipment to record yet.

bash
SHIP_ARTIFACT=$(~/.claude/skills/nanostack/bin/resolve.sh compound 2>/dev/null | jq -r '.upstream_artifacts.ship // empty')
SHIP_RUN_MODE=$(jq -r '.run_mode // "normal"' "$SHIP_ARTIFACT" 2>/dev/null || echo "normal")
2. Identify what's worth capturing

Not everything needs a solution document. Capture:

  • Bugs that took more than a trivial fix (the investigation is the value)
  • Patterns you want the agent to follow in future sprints
  • Architecture decisions with trade-offs that someone might question later

Skip:

  • Typos, formatting, trivial fixes
  • Standard library usage (the docs are better)
  • Findings that were auto-fixed with no investigation
3. Check for existing solutions

Before creating a new document, search for related ones:

bash
~/.claude/skills/nanostack/bin/find-solution.sh "relevant keywords"
~/.claude/skills/nanostack/bin/find-solution.sh --tag relevant-tag
~/.claude/skills/nanostack/bin/find-solution.sh --file affected/file/path

If a closely related solution exists:

  • Update it if the new information extends or corrects the existing document
  • Create a new one if it's a different problem that happens to share keywords

Do not create duplicates. One good document beats two partial ones.

When updating an existing solution, evolve it:

  1. Rewrite the compiled truth. Read the current Problem, Solution, and Prevention sections. Do they reflect what you know NOW, or what you knew when the document was created? If the new sprint changed your understanding, rewrite these sections completely — don't append to them. The compiled truth must always reflect the current best understanding, not a history of partial fixes.
  2. Increment applied_count in the frontmatter
  3. Set validated: true and last_validated to today's date
  4. Adjust confidence (1-10 scale): if the solution worked perfectly this sprint → set to current + 2 (cap 10). If it worked with adjustments → keep current. If it failed or needed major changes → set to current - 2 (floor 1). Confidence determines search ranking: high-confidence solutions surface first.
  5. Append a ### YYYY-MM-DD — Context entry to the ## History section at the bottom, documenting what happened in this sprint and what changed

Compiled truth is rewritten. Timeline is appended. The sections above History always reflect the latest understanding. The History section is the immutable evidence trail of how that understanding evolved. If old information was wrong, rewrite the compiled truth — don't leave stale text. The History entry records what changed and why.

Show full SKILL.md (369 more words)Show less
4. Write solution documents

For each significant learning, create a document:

bash
~/.claude/skills/nanostack/bin/save-solution.sh <type> "<title>" "tag1,tag2,tag3"

Types:

  • bug - a problem you encountered and solved
  • pattern - a recurring approach worth remembering
  • decision - an architecture or design choice with rationale

The script creates the file with YAML frontmatter and section templates. Fill in every section. Be specific:

Good:

markdown
## Problem
Stripe webhook endpoint accepted POST without signature verification.
stripe.webhooks.constructEvent() requires the raw request body, not parsed JSON.

## Solution
Use express.raw() middleware on the webhook route before express.json() parses it.

Bad:

markdown
## Problem
Webhook was broken.

## Solution
Fixed the webhook handler.

The value is in the detail. A future agent reading this needs enough context to apply the solution without re-investigating.

5. Update frontmatter

After filling in the body, update the frontmatter:

  • files: add the actual file paths involved
  • severity: adjust based on actual impact (critical, high, medium, low)
  • tags: add any tags that would help future search
6. Report

Print a summary of what was captured:

Compound: 3 solutions captured

  bug/stripe-webhook-signature.md (high) - Stripe webhook missing signature verification
  pattern/api-error-handling.md (medium) - Structured error responses with codes
  decision/auth-clerk-over-custom.md (medium) - Chose Clerk over custom auth

Total solutions in project: 12

After reporting, save the artifact. Run this command now — do not skip it:

bash
~/.claude/skills/nanostack/bin/save-artifact.sh compound '<json with phase, summary including solutions_created, solutions_updated, total_solutions, context_checkpoint including summary, key_files, decisions_made, open_questions>'
7. Check graduation candidates

After saving solutions, check if any existing solutions now meet graduation criteria:

bash
~/.claude/skills/nanostack/bin/graduate.sh

If candidates exist, report them:

N solutions ready to graduate into skill files.
Run `graduate.sh --apply` to promote them, or review with `graduate.sh` (dry-run).

Graduation means a proven, validated solution gets baked directly into a SKILL.md file — no more runtime lookup needed. Only solutions with 3+ applications, validated status, and existing referenced files qualify.

8. Check stale diarizations

If .nanostack/know-how/diarizations/ exists, check whether this sprint touched files covered by any existing diarization. If so, the diarization may be stale — suggest re-running it:

Diarization for <subject> is N days old and this sprint modified files it covers.
Consider re-running: /investigate --diarize <subject>

Then tell the user:

Knowledge captured. These solutions will be found automatically by /nano during planning and /review during code review.

Telemetry finalize

Before returning control:

bash
_F="$HOME/.claude/skills/nanostack/bin/lib/skill-finalize.sh"
[ -f "$_F" ] && . "$_F" compound success
unset _F

Pass abort or error instead of success if compound did not complete normally.

Rules

  • One problem per document. Don't combine unrelated fixes into one solution.
  • Fill every section. Empty sections are noise. If "What didn't work" is empty, either you fixed it on the first try (rare, skip the document) or you forgot to write it down.
  • Use the exact file paths. src/api/webhooks/stripe.ts is searchable. "The webhook file" is not.
  • Tags are for search, not decoration. Use terms someone would grep for: stripe, webhooks, hmac, not payment-processing-integration.
  • Set severity accurately. Solutions are ranked by severity when searched. Don't leave everything as medium.
  • Update, don't duplicate. If ~/.claude/skills/nanostack/bin/find-solution.sh returns a close match, update that document.
  • The Prevention section is the highest-value section. A bug fix helps once. A prevention rule helps every future sprint.

© garagon, Apache-2.0. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

Just SKILL.md in compound of garagon/nanostack.

Open the folder on GitHubat commit 0372aed

Compare with similar skills

Compound next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Compound compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Compound this skillgaragon/nanostack207—~2.2kAutomated safety check: PassApache-2.0
Learning to Learn (OpenMAIC)THU-MAIC/OpenMAIC40k—~502Automated safety check: PassMIT
Compound Learningsparcadei/Continuous-Claude-v33.9k1 repos~1.6kAutomated safety check: NotesMIT
Compound Learning WriterEveryInc/compound-engineering-plugin25k—~2kAutomated safety check: PassMIT
Compound Learnings RefreshEveryInc/compound-engineering-plugin25k—~2kAutomated safety check: PassMIT
Scikit LearnK-Dense-AI/scientific-agent-skills48k1 repos~3.3kAutomated safety check: NotesBSD-3-Clause

Similar skills

  • A Chinese-language skill that embeds learning strategies like retrieval practice and self-explanation as a parallel goal inside an OpenMAIC subject lesson, without making study skills the topic.

    40k GitHub stars~502 tokensUpdated today
    EducationAuto-check passed
  • Compound Learnings

    parcadei/Continuous-Claude-v3

    Transform session learnings into permanent capabilities (skills, rules, agents).

    3.9k GitHub starsUsed in 1 repo~1.6k tokens
    Auto-check: notes
  • Compound Learning Writer

    EveryInc/compound-engineering-plugin

    Records one solved and verified problem as a durable learning in the repository, but only when the reasoning is not already clear from the final code, tests or docs.

    25k GitHub stars~2k tokensUpdated today
    Agent WorkflowsAuto-check passed
  • Compound Learnings Refresh

    EveryInc/compound-engineering-plugin

    Audits a repo's stored learnings against the current codebase, fixes stale, overlapping or superseded docs and reports on every document.

    25k GitHub stars~2k tokensUpdated today
    Agent WorkflowsAuto-check passed
  • Scikit Learn

    K-Dense-AI/scientific-agent-skills

    Supports machine learning in Python with scikit-learn. An agent skill from K-Dense-AI/scientific-agent-skills.

    48k GitHub starsUsed in 1 repo~3.3k tokens
    Data & AnalyticsAuto-check: notes
  • Lets you review, search, prune and export the learnings gstack has collected across sessions, and surfaces them when a past fix or pattern comes up.

    136k GitHub stars~8.2k tokensUpdated today
    Agent WorkflowsAuto-check: notes

More from garagon/nanostack

All 14 skills in this repo
  • Nano

    garagon/nanostack

    A skill your agent uses when starting non-trivial work (touching 3+ files, new features, refactors, bug investigations).

    207 GitHub stars~3.3k tokensUpdated 27 days ago
    Auto-check passed
  • Nano Run

    garagon/nanostack

    First-time setup and guided sprint. An agent skill from garagon/nanostack.

    207 GitHub stars~3k tokensUpdated 27 days ago
    Auto-check passed
  • Security

    garagon/nanostack

    Use before shipping to production. An agent skill from garagon/nanostack.

    207 GitHub stars~3.7k tokensUpdated 27 days ago
    Auto-check: notes
  • Ship

    garagon/nanostack

    A skill your agent uses when code is ready to ship — creates PRs, merges, deploys, and verifies.

    207 GitHub stars~4.2k tokensUpdated 27 days ago
    Auto-check passed
  • Conductor

    garagon/nanostack

    Orchestrate parallel agent sessions through a sprint. An agent skill from garagon/nanostack.

    207 GitHub stars~2.6k tokensUpdated 27 days ago
    Auto-check passed
  • Feature

    garagon/nanostack

    Add a feature to an existing project with a full sprint. An agent skill from garagon/nanostack.

    207 GitHub stars~1.4k tokensUpdated 27 days ago
    Auto-check passed

Questions about Compound

What does Compound do?

Document what you learned during this sprint. An agent skill from garagon/nanostack. Compound is an agent skill from garagon/nanostack. Document what you learned during this sprint.

How do I install Compound in Claude Code?

Run `npx skills add garagon/nanostack --skill compound -a claude-code`. Or copy the skill folder (compound in garagon/nanostack) into .claude/skills/compound in your project. Claude Code loads it when a task matches its description.

How do I install Compound in Codex?

Run `npx skills add garagon/nanostack --skill compound -a codex`. Or copy the skill folder (compound in garagon/nanostack) into .agents/skills/compound in your project. Codex loads it when a task matches its description.

Can I use Compound in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add garagon/nanostack --skill compound -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/compound, .gemini/skills/compound, .github/skills/compound and .opencode/skills/compound in your project.

What does Compound need to run?

Going by SKILL.md and its folder, Compound needs the command-line tools its instructions call (jq).

Does Compound access the network?

SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.

Is Compound safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.

What licence does Compound use?

Compound is published under the Apache-2.0 licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Compound use?

About 2.2k tokens (SKILL.md is roughly 9k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.

What are the alternatives to Compound?

Skills that share tags, products or a category with Compound: Learning to Learn (OpenMAIC) (THU-MAIC/OpenMAIC, 40k stars), Compound Learnings (parcadei/Continuous-Claude-v3, 3.9k stars), Compound Learning Writer (EveryInc/compound-engineering-plugin, 25k stars) and Compound Learnings Refresh (EveryInc/compound-engineering-plugin, 25k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Compound?

garagon (a GitHub user) maintains it in garagon/nanostack, which has 207 GitHub stars. The repository holds 14 skills in this directory. The repository was last updated on September 10, 2026.

Source: garagon/nanostack on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.