The Agent Skills Directory

[COMMAND_EXECUTION]: Executes shell commands via the Bash tool and uses the fh-run CLI to manage agent dispatches.
[EXTERNAL_DOWNLOADS]: Fetches and executes the @chrono-meta/fh-gate package from the npm registry using the npx utility (SKILL.md).
[DYNAMIC_EXECUTION]: Dynamically assembles agent briefs, context cards, and execution plans at runtime, which are then processed via the Agent tool or the fh-run command.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: Reads capability definitions and project context from files such as .claude/agents/*.md, LOCAL_SKILL_REGISTRY.md, .mcp.json, and CLAUDE.md (SKILL.md, SKILL_detail.md).
Boundary markers: Utilizes Session Context Cards and a token-limited Agent Registry to provide context to isolated sub-agents.
Capability inventory: The skill has access to file system reading (Read, Glob, Grep), shell execution (Bash), and agent/skill invocation (Agent, Skill).
Sanitization: No specific sanitization or validation logic is defined for the content extracted from local registry or configuration files.

agent-composer