The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides commands to install additional child skills using pnpm dlx skills@latest. These resources are fetched from the skill author's repository (patinaproject/skills) and a third-party developer (mattpocock/skills). The commands specifically include npm_config_ignore_scripts=true, which follows security best practices by preventing the execution of potentially malicious lifecycle scripts during the installation process.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it ingests untrusted data from bug report links and operator instructions.
Ingestion points: Data enters the context via the <report-link> and additional operator instructions in the /fix command.
Boundary markers: The skill does not explicitly define delimiters or instructions to ignore embedded commands within the bug reports.
Capability inventory: The skill has the capability to modify code, run tests, and publish pull requests through child skills like implement, tdd, and ready-pr (SKILL.md).
Sanitization: There is no mention of sanitization or filtering for the external report content. However, the risk is mitigated by the structured workflow requiring red-to-green verification on a deployed head and the use of the polish quality loop.

fix