The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/baseline.mts implements a self-building mechanism that uses bun build to compile its source code into a bundle within a .build directory at runtime before executing it.
[COMMAND_EXECUTION]: The orchestration scripts (sweep.mts, probe.mts, ledger.mts, baseline.mts) utilize system command execution via spawnSync and execFile to invoke tools such as git, ps, and lsof for monitoring worktrees and process activity.
[EXTERNAL_DOWNLOADS]: The scripts/baseline.mts script executes bun install during its build phase to fetch the diff package from a registry at runtime.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from the repository, including code changes and commit messages, as part of its lead orchestration role.
Ingestion points: File contents, git history, and subagent completion signals are ingested through scripts/sweep.mts and scripts/probe.mts as described in Section 2 of SKILL.md.
Boundary markers: The instructions mandate human review gates and reference external review skills for verification.
Capability inventory: The skill possesses capabilities for file system access, system command execution, and git-based network operations.
Sanitization: The skill implements a PII scanning feature that searches for sensitive identifiers using a user-defined 'needle' file before changes are landed on the mainline branch (SKILL.md, Section 6).

orchestrator-mode