The Agent Skills Directory

[NO_CODE]: The skill consists entirely of natural language instructions and a markdown template for manual reporting. It does not include any scripts, executable code, or package dependencies.
[SAFE]: The instructions include explicit security constraints, such as 'Read-only. Do not edit files automatically,' which prevents the agent from making unauthorized changes to the filesystem.
[INDIRECT_PROMPT_INJECTION]: The skill defines a surface for processing external context (project decisions), which could theoretically contain malicious instructions. However, because the skill has no write capabilities, network access, or dynamic execution paths, the risk of exploitation is negligible.

decision-capture