The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill utilizes npx @claude-flow/cli@latest, which downloads and executes external code at runtime from the NPM registry. This package is not from a verified trusted organization.
[COMMAND_EXECUTION]: The workflow involves automated execution of various shell commands including npm test, eslint, and git based on the logic of the iterative loop.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection due to its processing of external command outputs.
Ingestion points: Ingests output from npm test, git log, and file contents into the agent's context (e.g., in the 'Iteration pattern' steps).
Boundary markers: Missing specific delimiters or instructions to treat ingested data as untrusted content to prevent the agent from obeying instructions embedded in logs or files.
Capability inventory: Executes shell commands and high-level agent tasks via claude_execute.
Sanitization: No evidence of sanitization or filtering of the ingested data before it influences subsequent loop iterations.

Iterative Loop