clean-learn
Turn confirmed defects, false positives, accepted changes, review outcomes, and benchmark results into reversible policy, threshold, adapter, convention, or…
Skill by shashank-sn
Chalk Labs
Need help deploying clean-learn?
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/shashank-sn-clean-learn
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: security
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/clean-learn
What's inside
- Skill name
clean-learn- Description
- Turn confirmed defects, false positives, accepted changes, review outcomes, and benchmark results into reversible policy, threshold, adapter, convention, or calibration proposals. Use after an audit or repeated outcome reveals a useful local rule change; keep every proposal separate from execution and approval.
- License
- MIT
- Instructions
- 24 lines in SKILL.md · view on GitHub
Files (3)
- SKILL.md
- agent.json
- openai.yaml · agents/
SKILL.md (first lines)
# Clean Learn Propose narrow local improvements from confirmed outcomes. ## Workflow 1. Start from a confirmed outcome: escaped defect, caught defect, false positive, correct silence, repeated exception, or accepted local convention. 2. Link the source audit receipt, evidence hashes, affected rule or adapter, and observed consequence. 3. Propose one reversible change with scope, expected benefit, risks, migration, rollback, and a calibration fixture. 4. For a bottom-up proposal, preserve at least two distinct supporting evidence hashes, a distinct clean control, an untouched held-out fixture with passing evidence, the expected false-positive cost, and rollback. 5. Compare the proposal with the currently trusted policy and flag every new command, permission, suppression, threshold change, and gate-status change. 6. Send the proposal to a separate reviewer. Apply it only after explicit approval from a human who is independent of both proposer and reviewer. 7. Run existing and new calibration fixtures, then record measured detection and false-positive outcomes. 8. Run `clean-code learn --proposal <proposal.json>` before sending the proposal for approval. ## Hard boundaries - Never let the proposing agent approve or activate its proposal. - Never suppress correctness, safety, security, privacy, data-integrity, or explicit requirement failures. - Never turn missing, unavailable, stale, or unrun evidence into success. - Never generalize from one stylistic preference into a universal rule. - Keep rollback data and provenance for every accepted policy change.
From shashank-sn/clean-code, MIT.
More from shashank-sn/clean-code
- 0.523score
clean-audit
shashank-sn
Produce an immutable, revision-bound release receipt from deterministic verification, independent test planning, review, human spot checks, policy identity…
SkillBFreshness: BClaude Code +9 - 0.523score
clean-build
shashank-sn
Implement one bounded behavior through short, verified steps while protecting requirements, architecture policy, repository conventions, and unrelated code.
SkillBFreshness: BClaude Code +9 - 0.523score
clean-lfg
shashank-sn
Run the full autonomous shipping pipeline from planning through implementation, simplification, dead-code pruning, final verification, review, ship, and CI…
SkillBFreshness: BClaude Code +9 - 0.523score
clean-merge-resolver
shashank-sn
Resolve an in-progress git merge or rebase conflict hunk by hunk, tracing each resolution to both sides' primary sources, then finish the operation.
SkillBFreshness: BClaude Code +9 - 0.523score
clean-orchestrate
shashank-sn
Coordinate specification, implementation, independent testing, deterministic verification, review, human spot checks, and audit without treating agent…
SkillBFreshness: BClaude Code +9 - 0.523score
clean-review
shashank-sn
Review code, requirements, architecture, tests, and deterministic evidence with concrete findings, structural simplification, and permission to return zero…
SkillBFreshness: BClaude Code +9 - 0.523score
clean-eval-discover
shashank-sn
Turn real work outcomes and bounded human judgment into representative, reviewable evaluation candidates.
SkillBFreshness: BClaude Code +9 - 0.523score
clean-design
shashank-sn
Turn requirements into use cases, stable policy boundaries, delivery and data boundaries, dependency direction, acceptance examples, and repository-owned…
SkillBFreshness: BClaude Code +9 - 0.523score
clean-refactor
shashank-sn
Improve names, functions, duplication, responsibility placement, dependency direction, boundaries, and module structure through behavior-preserving changes.
SkillBFreshness: BClaude Code +9 - 0.520score
clean-discover
shashank-sn
Inspect a repository read-only to identify languages, project metadata, configuration, and available paths for deterministic quality checks.
SkillBFreshness: BClaude Code +9 - 0.520score
clean-auditor
shashank-sn
Produce the immutable release receipt mapping requirements to evidence, decoupled from implementers so it cannot self-whitewash.
SkillBFreshness: BClaude Code +9 - 0.520score
clean-brainstorm
shashank-sn
Explore vague product ideas into a requirements-only plan with scope boundaries, acceptance signals, and explicit non-goals.
SkillBFreshness: BClaude Code +9
Signals
- Stars
- 0
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Aug 24, 2026 (2 months ago)
- Last release
- Sep 17, 2026 (1 month ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/shashank-sn/clean-code
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from shashank-sn
- 0.607score
hyv-prepare-rebuild
shashank-sn
Prepare a rebuild task only after an upstream REBUILD recommendation, a CopySpec, and a signed rebuild-authorization capability.
SkillBFreshness: B★ 32Claude Code +9 - 0.604score
hyv-apply-hidden-text-policy
shashank-sn
Apply only explicitly approved minimal hidden-text removals and return hashes, exact changes, and remaining findings.
SkillBFreshness: B★ 32Claude Code +9 - 0.604score
hyv-learning
shashank-sn
Inspect or explicitly mutate text-free local learning state for a profile.
SkillBFreshness: B★ 32Claude Code +9 - 0.604score
hyv-analyze
shashank-sn
Run separate VoiceDNA and AI Editor checks plus a non-scoring Unicode hygiene inspection against a draft.
SkillBFreshness: B★ 32Claude Code +9 - 0.604score
hyv-logic-lint
shashank-sn
Run the deterministic document-coherence gate for topic drift, unanchored inference, and internal contradictions.
SkillBFreshness: B★ 32Claude Code +9 - 0.604score
hold-your-voice
shashank-sn
Review or revise user-provided writing while protecting its stated meaning, constraints, and voice.
SkillBFreshness: B★ 32Claude 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 clean-learn 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
clean-learn skill. Agent Mart. https://agentmart.dev/skills/shashank-sn-clean-learn (last updated 2026-10-24)