improve
Self-improvement — analyze the project and update AGENTS.md, rules, settings, and agents to improve AI effectiveness
Skill by BostonOrange
Chalk Labs
Need help deploying improve?
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/bostonorange-improve
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
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/improve
What's inside
- Skill name
improve- Description
- Self-improvement — analyze the project and update AGENTS.md, rules, settings, and agents to improve AI effectiveness
- License
- MIT
- Instructions
- 154 lines in SKILL.md · view on GitHub
Files (1)
- SKILL.md
SKILL.md (first lines)
# Improve — Framework Self-Improvement
Analyze the current project state and improve the .codex/ configuration for better AI assistance.
## Usage
```
/improve — Full improvement pass (all areas)
/improve agents-md — Update AGENTS.md only (fill placeholders, add patterns)
/improve rules — Update rule file patterns to match actual project
/improve settings — Update permissions and model config
/improve agents — Tune agent tools and models for project needs
/improve scan — Report only, no changes (dry run)
```
## Process
### Phase 1: Project Discovery
Scan the project to build a comprehensive profile:
1. **Tech stack detection:**
```bash
# Package managers and dependency files
ls package.json requirements.txt Pipfile go.mod Cargo.toml build.gradle pom.xml Gemfile composer.json sfdx-project.json 2>/dev/null
```
2. **File type census:**
```bash
find . -type f -not -path "*/.git/*" -not -path "*/node_modules/*" -not -path "*/__pycache__/*" -not -path "*/vendor/*" -not -path "*/.next/*" -not -path "*/dist/*" -not -path "*/build/*" | sed 's/.*\.//' | sort | uniq -c | sort -rn | head -20
```
3. **Framework detection:** Read config files (next.config.*, vite.config.*, angular.json, etc.)
4. **Directory structure:** Map the top-level architecture
5. **Existing tooling:** Find linter configs, test configs, CI configs
### Phase 2: AGENTS.md Improvement
Read current AGENTS.md and check:
1. **Unfilled placeholders:** Find any remaining `{{...}}` and fill from discovered data:
- `{{PROJECT_DESCRIPTION}}` — from README.md, package.json description, etc.
- `{{TECH_STACK_TABLE}}` — build from dependency files
- `{{CODE_STRUCTURE}}` — generate directory tree
- `{{CODING_STANDARDS}}` — infer from linter configs (.eslintrc, .prettierrc, pyproject.toml, etc.)
- `{{ERROR_HANDLING_PATTERN}}` — find common error patterns in code
- `{{TESTING_STRATEGY}}` — infer from test config and existing tests
- `{{INTEGRATIONS}}` — list discovered external services
2. **Missing information:** Add sections for:
- Discovered environment variables and their purposes
- Common commands found in package.json scripts or Makefile
- Team conventions visible in git log patterns
3. **Outdated information:** Check if documented commands, paths, or patterns still exist
If `--scope agents-md` or full pass: apply changes. Otherwise skip.
### Phase 3: Rules Improvement
For each rule file in `.codex/rules/`:
1. **Verify patterns match reality:**
- Extract the glob patterns from the YAML frontmatter
- Check if files matching those patterns actually exist
- If patterns are still `{{PLACEHOLDER}}`, replace with discovered file patterns
2. **Add missing rules:**
- If the project has file types not covered by any rule, consider adding rules
- Common gaps: middleware files, config schemas, migration files, test utilities
3. **Remove irrelevant rules:**
- If a rule's file patterns match zero files, warn (don't delete — user may add files later)
If `--scope rules` or full pass: apply changes. Otherwise skip.
### Phase 4: Settings Improvement
Read `.codex/settings.local.json` and check:
1. **Permissions:** Are all tools the project workflow needs allowed?
- If project uses git heavily, ensure Bash(git*) patterns
- If project needs web access for docs, ensure WebFetch/WebSearch
2. **Model:** Is the default model appropriate?
- Complex architecture projects may benefit from opus
- High-volume code generation may benefit from sonnet for speed
If `--scope settings` or full pass: apply changes. Otherwise skip.
### Phase 5: Agent Improvement
For each agent in `.codex/agents/`:
1. **Tool appropriateness:** Do the allowed tools match what the agent needs?
2. **Model selection:** Is the model appropriate for the task complexity?
3. **Placeholder check:** Are any `{{PLACEHOLDER}}` values unfilled?
If `--scope agents` or full pass: apply changes. Otherwise skip.
### Phase 6: Report
```
## Framework Improvement Report
### Project Profile
- **Type:** {detected project type}
- **Stack:** {tech stack summary}
- **Size:** {file count, LOC estimate}
### Changes Made
| File | Change | Reason |
|------|--------|--------|
| {path} | {description} | {why} |
### AGENTS.md Status
- Placeholders filled: {n}From BostonOrange/claude-code-framework, MIT.
More from BostonOrange/claude-code-framework
- 0.514score
ccf-architecture-contract
BostonOrange
Load when you need to know WHY claude-code-framework is shaped the way it is before changing its structure — questions like "where does this new file belong…
SkillCFreshness: C★ 3Claude Code +9 - 0.514score
ccf-parity-playbook
BostonOrange
Load before touching setup.sh or setup.ps1 in claude-code-framework, or when asked "does the PowerShell installer match the bash one?", "why does setup work…
SkillCFreshness: C★ 3Claude Code +9 - 0.511score
install-framework
BostonOrange
Install the Claude Code Framework into the current repo by cloning it from GitHub (no local clone path needed).
SkillCFreshness: C★ 3Claude Code +9 - 0.510score
ccf-doc-sync-campaign
BostonOrange
Load when a count or table in claude-code-framework docs looks stale, when adding/removing a skill/agent/command/rule/hook (which changes counts everywhere)…
SkillCFreshness: C★ 3Claude Code +9 - 0.510score
ccf-build-and-env
BostonOrange
Load when setting up a dev environment for the claude-code-framework repo itself — fresh clone, new machine, new agent session — or when asked "how do I…
SkillCFreshness: C★ 3Claude Code +9 - 0.510score
ccf-dogfood-and-drift
BostonOrange
Load when working with claude-code-framework's own .claude/ directory or when tests/check-dogfood-drift.sh fails — symptoms like "Unexpected .claude/template…
SkillCFreshness: C★ 3Claude Code +9 - 0.510score
ccf-onboarding-tour
BostonOrange
Load this at the start of any first session in the claude-code-framework repo — when you (human or model) have zero context and need the right mental model…
SkillCFreshness: C★ 3Claude Code +9 - 0.510score
ccf-release-and-install
BostonOrange
Load when installing or uninstalling the claude-code-framework in a target repo, or answering how installation works — driving setup.sh / setup.ps1, choosing…
SkillCFreshness: C★ 3Claude Code +9 - 0.508score
team
BostonOrange
Spawn pre-configured agent teams for parallel review, analysis, or implementation tasks
SkillCFreshness: C★ 3Claude Code +9 - 0.489score
ccf-failure-archaeology
BostonOrange
Load before re-investigating anything in claude-code-framework that smells like a past battle — "why did the PowerShell CI job fail?", "didn't we already fix…
SkillCFreshness: C★ 3Claude Code +9 - 0.485score
ccf-testing-and-qa
BostonOrange
Load this before claiming any claude-code-framework change is "done", before running or extending tests/run-all.sh, or when a check-*.sh test fails and you…
SkillCFreshness: C★ 3Claude Code +9 - 0.485score
ccf-change-control
BostonOrange
Load before committing, merging, or ending any session that modified framework files in claude-code-framework.
SkillCFreshness: C★ 3Claude Code +9
Signals
- Stars
- 3
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Jun 4, 2026 (4 months 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 BostonOrange
- 0.514score
ccf-parity-playbook
BostonOrange
Load before touching setup.sh or setup.ps1 in claude-code-framework, or when asked "does the PowerShell installer match the bash one?", "why does setup work…
SkillCFreshness: C★ 3Claude Code +9 - 0.514score
ccf-architecture-contract
BostonOrange
Load when you need to know WHY claude-code-framework is shaped the way it is before changing its structure — questions like "where does this new file belong…
SkillCFreshness: C★ 3Claude Code +9 - 0.511score
install-framework
BostonOrange
Install the Claude Code Framework into the current repo by cloning it from GitHub (no local clone path needed).
SkillCFreshness: C★ 3Claude Code +9 - 0.510score
ccf-dogfood-and-drift
BostonOrange
Load when working with claude-code-framework's own .claude/ directory or when tests/check-dogfood-drift.sh fails — symptoms like "Unexpected .claude/template…
SkillCFreshness: C★ 3Claude Code +9 - 0.510score
ccf-doc-sync-campaign
BostonOrange
Load when a count or table in claude-code-framework docs looks stale, when adding/removing a skill/agent/command/rule/hook (which changes counts everywhere)…
SkillCFreshness: C★ 3Claude Code +9 - 0.510score
ccf-build-and-env
BostonOrange
Load when setting up a dev environment for the claude-code-framework repo itself — fresh clone, new machine, new agent session — or when asked "how do I…
SkillCFreshness: C★ 3Claude 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 improve 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
improve skill. Agent Mart. https://agentmart.dev/skills/bostonorange-improve (last updated 2026-10-24)