claude-skills
AGENTS.md0 starsChanged 9 months agoMITView on GitHub
Agent-facing project context. Shared across Claude Code and Codex CLI for cross-agent continuity. The same 4 agents are available as Codex custom roles, each defined in a per-file role TOML under ~/.codex/agents/<name>.toml (personal) or .codex/agents/<name>.toml (project-level). The inline [agents.<name>] form in config.toml is a legacy alternate with known config-loading caveats (openai/codex#14579, #15250) — prefer the per-file files below: Enable with multi_agent = true in [features]. See forge-codex-multiagent SKILL.md Section 1b for config syntax. See README.md for the full skill catalog…
In this repository17 other files it gives its agents.
- auditing-skillsSkill
- creating-skillsSkill
- forge-builderSkill
- forge-claude-teamsSkill
- forge-codex-multiagentSkill
- forge-orchestratorSkill
- forge-researchSkill
- handoff-freshSkill
- handoffSkill
- managing-doc-manifestSkill
- okfSkill
- pickupSkill
- studySkill
- syncing-docsSkill
- testing-skillsSkill
- updating-skillsSkill
- wrapSkill
The file
Files
Agents
Agent-facing project context. Shared across Claude Code and Codex CLI for cross-agent continuity.
Forge Skills (Autonomous Workflows)
Skills
| Skill | Layer | Purpose |
| forge-claude-teams | 0 (practices) | Claude Code best practices for Agent tool, TeamCreate, SendMessage, task coordination |
| forge-codex-multiagent | 0 (practices) | Codex CLI best practices for spawn_agent, fork_context, batch processing |
| forge-research | 1 (capability) | Autonomous multi-agent research with hypothesis tracking and adversarial challenge |
| forge-builder | 1 (capability) | Autonomous building/coding with self-review, self-improvement, and quality gates |
| forge-orchestrator | 2 (orchestrator) | Sequences Layer 1 skills through milestone-gated cycles with review and compound learning |
Custom Agents
| Agent | Purpose | Tools |
| forge-adversarial-reviewer | Critical review with confidence-gated findings (>80%) | Read, Grep, Glob, Bash |
| forge-build-worker | Implementation worker with file-scope ownership | Read, Write, Edit, Bash, Grep, Glob |
| forge-research-worker | Web research and codebase exploration (read-only, Rule of Two) | Read, Grep, Glob, Bash, WebSearch, WebFetch |
| forge-performance-auditor | Metric-driven performance auditing | Read, Bash, Grep, Glob |
Codex CLI Agent Roles
The same 4 agents are available as Codex custom roles, each defined in a per-file role TOML under ~/.codex/agents/<name>.toml (personal) or .codex/agents/<name>.toml (project-level). The inline [agents.<name>] form in config.toml is a legacy alternate with known config-loading caveats (openai/codex#14579, #15250) — prefer the per-file files below:
| Role | Config File | Sandbox |
forge-adversarial-reviewer | .codex/agents/forge-adversarial-reviewer.toml | read-only |
forge-build-worker | .codex/agents/forge-build-worker.toml | danger-full-access |
forge-research-worker | .codex/agents/forge-research-worker.toml | read-only |
forge-performance-auditor | .codex/agents/forge-performance-auditor.toml | read-only |
Enable with multi_agent = true in [features]. See forge-codex-multiagent SKILL.md Section 1b for config syntax.
Cross-References
- forge-research and forge-builder read the platform practices guide on startup: forge-claude-teams (Claude Code) or forge-codex-multiagent (Codex CLI)
- forge-orchestrator invokes forge-research for unknowns and forge-builder for implementation
- forge-orchestrator spawns all 4 custom agents for review, building, research, and performance auditing
Dependency Graph
forge-orchestrator (Layer 2)
├── READS: forge-claude-teams OR forge-codex-multiagent (Layer 0)
├── INVOKES: forge-research (Layer 1)
├── INVOKES: forge-builder (Layer 1)
└── SPAWNS: forge-adversarial-reviewer, forge-build-worker,
forge-research-worker, forge-performance-auditor
forge-research (Layer 1)
├── READS: forge-claude-teams OR forge-codex-multiagent (Layer 0)
└── SPAWNS: forge-adversarial-reviewer (optional)
forge-builder (Layer 1)
├── READS: forge-claude-teams OR forge-codex-multiagent (Layer 0)
├── SPAWNS: forge-build-worker (optional)
└── SPAWNS: forge-adversarial-reviewer (optional)Existing Skills
See README.md for the full skill catalog and installation instructions.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.