mawf-compound
Capture a solved-and-verified problem as a durable solution doc in the MAWF knowledge store.
Skill by imBlanker
Chalk Labs
Need help deploying mawf-compound?
Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/imblanker-mawf-compound
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Use cases: documents
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/mawf-compound
What's inside
- Skill name
mawf-compound- Description
- Capture a solved-and-verified problem as a durable solution doc in the MAWF knowledge store. Use only after verified work produced non-obvious reasoning absent from final code, tests, and existing docs.
- License
- MIT
- Instructions
- 65 lines in SKILL.md · view on GitHub
Files (3)
- SKILL.md
- frontmatter-schema.md · references/
- grounding-validation.md · references/
SKILL.md (first lines)
# MAWF Solution Capture
Adapted from EveryInc compound-engineering-plugin (MIT), © Every — see `NOTICE-MAWF-INTEGRATIONS.md` at the skills root. Storage, retrieval, and validation run through the `mawf knowledge` CLI.
## 1. Hard gate
Document only a problem that is **solved and verified**. Then apply the durable bar, judged by this counterfactual: if the learning document disappeared, would a future engineer reading the final implementation still be likely to repeat the mistake or redo substantial investigation? A learning earns its place only when it is:
- **non-obvious** — not readily recoverable from the final code, tests, types, comments, or existing docs;
- **durable** — project reasoning that persists;
- **material** — losing it would plausibly cause recurrence, material risk, or substantial rediscovery.
All three must hold. Additional rules:
- **One learning per run.** A session that produced several gets several sequential runs, never one batched run.
- Completion phrases ("that worked", "it's fixed", "problem solved") identify the checkpoint but do NOT establish eligibility.
- Effort, diff size, and "it's done" never qualify a capture.
- If the counterfactual fails: write nothing and report why, judged from the session rather than by asking.
## 2. Dedup before creating
Search first: `mawf knowledge search <keywords>` (and `mawf knowledge context` for injection). Score overlap across five dimensions: problem statement, root cause, solution approach, referenced files, prevention rules.
| Overlap | Action |
| --- | --- |
| High (4–5 dimensions) | **Update the existing doc in place**: keep its path and structure; add `last_updated: YYYY-MM-DD` to the frontmatter; do not retitle unless the framing shifted |
| Moderate (2–3) | Create a new doc AND flag both docs for consolidation review |
| Low / none (0–1) | Create normally |
An existing learning that became materially inaccurate or incomplete qualifies for capture — update that learning instead of creating a duplicate that leaves the stale one misleading.
## 3. Document format
Path: `docs/knowledge/solutions/<category>/yyyy-mm-dd-<slug>.md` (the date prefix is MAWF convention; the category comes from the existing taxonomy when one fits).
Frontmatter is CE-compatible YAML — required: `module`, `date`, `problem_type`, `component`, `severity`; optional: `symptoms`, `root_cause`, `resolution_type`, `tags` (≤8), `last_updated` (update-only). Body: `# Solution: <title>` header plus one `Status:` line — `Status: candidate` until evidence is complete, `Status: verified` after, `Status: superseded` when replaced.
Evidence-chain sections (canonical English headings): Problem / Symptom / Root cause / Failed attempts / Fix / Why it works / Prevention / Scope / Evidence / Alternatives considered. Which sections are machine-required per status is enforced by `mawf knowledge verify` (candidates need at minimum Problem, Symptom, Fix, Alternatives). Field enums, corpus-first vocabulary, and YAML safety rules: [references/frontmatter-schema.md](references/frontmatter-schema.md).
## 4. Evidence rules
- Ground code-behavior claims in source, not conversation memory: read the defining line at the CURRENT tree and cite `file:line` alongside the claim.
- Merge-state claims cite PR numbers rather than bare commit SHAs; unmerged fixes are phrased as pending.
- A claim that cannot be verified against the tree is softened or attributed ("per this session's conclusion…") — never stated as fact. Never present unverified as verified.
- Candidates without complete evidence stay `Status: candidate`. A lightweight or low-budget run does not remove the evidence requirement; it only bounds how much investigation you may do.
## 5. Write boundary
Only the task's coordinator (or an explicitly named writer) writes the durable store; subagents write their own scratch only. Durable writes MUST go through the CLI so compare-and-swap, locking, sidecars, index invalidation and the archive seal all hold — direct `Write/Edit` to `docs/knowledge/**` bypasses them and is not sanctioned:
```bash
mawf knowledge create solution <category>/<yyyy-mm-dd>-<slug>.md --file <candidate.md> --source-task <task>
mawf knowledge update solution <relPath> --file <candidate.md> --expected-hash <hash> \
--verify-method <method> --verify-result pass # the ONLY way lastVerified moves
```
## 6. Before finishing
```bash
mawf knowledge verify
```
Claim flags from validation are adjudicated by you — fix, annotate, or confirm intentional; never an automatic rewrite or an automatic pass. Adjudication semantics: [references/grounding-validation.md](references/grounding-validation.md).From imBlanker/multi-agents-workflow, MIT.
More from imBlanker/multi-agents-workflow
- 0.549score
mawf-decision
imBlanker
Author and maintain long-lived decision notes in the MAWF knowledge store (write-notes semantics).
SkillBFreshness: B★ 1Claude Code +9 - 0.549score
mawf-archify
imBlanker
Render, validate, and deliver architecture diagrams through the MAWF-locked Archify engine.
SkillBFreshness: B★ 1Claude Code +9 - 0.547score
mawf-knowledge
imBlanker
Retrieve prior decisions and solutions from the MAWF knowledge store and inject bounded context.
SkillBFreshness: B★ 1Claude Code +9 - 0.543score
trellis-brainstorm
imBlanker
Guides collaborative requirements discovery before implementation.
SkillBFreshness: B★ 1Claude Code +9 - 0.541score
mawf-cost-guard
imBlanker
Check real-spend cost limits and concurrency before spawning agents or continuing a mawf workflow.
SkillBFreshness: B★ 1Claude Code +9 - 0.541score
mawf-loop
imBlanker
Run open-ended coding or research work with an iterative act, observe, reflect, and decide loop.
SkillBFreshness: B★ 1Claude Code +9 - 0.541score
mawf-orchestration
imBlanker
Plan and coordinate cost-bounded multi-agent workflows for complex projects using mawf.
SkillBFreshness: B★ 1Claude Code +9 - 0.541score
mawf-planner
imBlanker
Choose an agent architecture from project signals, host capabilities, risk, parallelism, and cost constraints.
SkillBFreshness: B★ 1Claude Code +9 - 0.541score
mawf-graph
imBlanker
Plan and inspect graph workflows with review gates, branching, checkpoints, and persistent execution state.
SkillBFreshness: B★ 1Claude Code +9 - 0.522score
domain-modeling
imBlanker
Build and sharpen a project's domain model.
SkillBFreshness: B★ 1Claude Code +9 - 0.522score
grilling
imBlanker
Grill the user relentlessly about a plan, decision, or idea.
SkillBFreshness: B★ 1Claude Code +9
Signals
- Stars
- 1
- Forks
- 0
- Open issues
- 2
- Last commit (this folder)
- Sep 17, 2026 (1 month ago)
- Last release
- Sep 17, 2026 (1 month ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from imBlanker
- 0.549score
mawf-archify
imBlanker
Render, validate, and deliver architecture diagrams through the MAWF-locked Archify engine.
SkillBFreshness: B★ 1Claude Code +9 - 0.549score
mawf-decision
imBlanker
Author and maintain long-lived decision notes in the MAWF knowledge store (write-notes semantics).
SkillBFreshness: B★ 1Claude Code +9 - 0.547score
mawf-knowledge
imBlanker
Retrieve prior decisions and solutions from the MAWF knowledge store and inject bounded context.
SkillBFreshness: B★ 1Claude Code +9 - 0.543score
trellis-brainstorm
imBlanker
Guides collaborative requirements discovery before implementation.
SkillBFreshness: B★ 1Claude Code +9 - 0.541score
mawf-graph
imBlanker
Plan and inspect graph workflows with review gates, branching, checkpoints, and persistent execution state.
SkillBFreshness: B★ 1Claude Code +9 - 0.541score
mawf-cost-guard
imBlanker
Check real-spend cost limits and concurrency before spawning agents or continuing a mawf workflow.
SkillBFreshness: B★ 1Claude Code +9
Done for you
Rather not set it up yourself?
Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.
See how we build AI agents →Read the case studies
Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.
- launches shipped
- 50+
- launches shipped
- media placements
- 500+
- media placements
- founders on every account
- 2
- founders on every account
Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block
Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.
Get mawf-compound running in your business
Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.
Cite this page
mawf-compound skill. Agent Mart. https://agentmart.dev/skills/imblanker-mawf-compound (last updated 2026-10-24)