The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires the installation of the claude-flow CLI from the npm registry as a prerequisite for use.
Evidence: npm install -g claude-flow@alpha in SKILL.md.
[COMMAND_EXECUTION]: The skill relies on shell commands for its core development workflows, including session management, git operations, and test execution.
Evidence: Commands such as claude-flow pair --start, $commit, $test, and $lint are central to the skill's operation as documented in SKILL.md.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted code from the local filesystem during pair programming sessions, which can lead to indirect prompt injection if that code contains malicious instructions.
Ingestion points: Untrusted code and data enter the agent context through commands like $review, $explain, $suggest, and $implement documented in SKILL.md.
Boundary markers: There are no explicit boundary markers or instructions to ignore embedded commands within the processed code.
Capability inventory: The skill possesses capabilities for file system modification ($implement, $refactor, $document), git operations ($commit, $branch), and arbitrary shell command execution through integrated testing and linting tools ($test, $lint).
Sanitization: No sanitization or validation mechanisms are described for the content being processed.

Pair Programming