The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill frequently executes remote code using npx with packages like ruv-swarm and claude-flow. These packages are not from verified or trusted sources and are invoked for core project management tasks (e.g., npx ruv-swarm github board-init and npx ruv-swarm github issue-to-swarm in SKILL.md).
[COMMAND_EXECUTION]: Extensive use of shell scripts and the GitHub CLI (gh) to perform operations like creating issues, editing project boards, and iterating through lists of items. The output of remote tools is used to control logic flow in shell scripts, such as the case statement driven by npx ruv-swarm github analyze-stale in the 'Stale Issue Management' section.
[INDIRECT_PROMPT_INJECTION]: The skill has a significant attack surface for indirect prompt injection as it processes data from an external platform.
Ingestion points: Reads GitHub issue content, labels, and user comments via commands like gh issue view and gh issue list in SKILL.md.
Boundary markers: No explicit boundary markers or 'ignore' instructions are provided to distinguish between legitimate issue content and potentially malicious embedded commands.
Capability inventory: The skill has capabilities to write back to GitHub (creating/editing issues, adding project items, posting comments) and execute external scripts via npx as documented in SKILL.md.
Sanitization: There is no visible sanitization or validation of retrieved issue bodies or comments before they are processed by the swarm agents or used in shell command loops.

qe-github-project-management