The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The scripts/bootstrap.ts file automatically runs bun install --frozen-lockfile to manage the skill's own dependencies. This targets the local package.json and uses the standard Bun package manager to fetch well-known libraries like commander.
[COMMAND_EXECUTION]: The skill frequently executes standard development tools including git, gh (GitHub CLI), gt (Graphite CLI), and bun. These executions are well-defined within the skill's scripts (github.ts, store.ts, worktree-audit.sh) to perform repository management tasks.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from GitHub pull request comments and automated review bots. However, it contains explicit security instructions in playbooks/babysit.md and references/bugbot-triage.md to treat this data as untrusted, verify claims against the code, and use safe execution patterns (like JSON payloads for gh api) to prevent shell injection or accidental instruction following.
[DYNAMIC_EXECUTION]: The framework uses Bun.spawnSync and child_process.spawn to run its internal utility scripts (e.g., watch-pr, orch). This is a legitimate architectural choice to provide the agent with enhanced repository interaction capabilities.

poteto-mode