The Agent Skills Directory

[Indirect Prompt Injection Surface]: The skill processes untrusted data from external planning issues and comment threads, which could potentially contain malicious instructions. This is a security consideration; however, the skill mandates that such content be treated strictly as data and includes explicit rules to ignore and flag injection attempts.\n
Ingestion points: Planning issue bodies and comment threads (SKILL.md).\n
Boundary markers: Includes specific 'Golden Rules' and instructions to treat external content as data only.\n
Capability inventory: The skill is limited to resolving metadata and generating command strings; it is explicitly forbidden from executing these commands itself.\n
Sanitization: Instructions require the agent to flag suspected injection attempts and proceed with the intended release flow.\n- [Local Command Execution]: The skill instructs the agent to execute a pre-flight validation command using a local Python module (setup_preflight) from the .apache-magpie-local directory. This is used for internal project consistency checks and environment verification.\n- [Sensitive Metadata Handling]: The skill resolves and utilizes sensitive metadata, such as Apache IDs and GPG fingerprints, to generate accurate release promotion commands. This data is used only to construct instructions for a human user to review and execute.

promote