The Agent Skills Directory

[COMMAND_EXECUTION]: The skill explicitly directs the agent in 'Step 2' and 'Step 6' to 'Actually run' setup, build, run, and test commands identified in the repository's documentation. This involves executing arbitrary shell commands sourced from untrusted local files.
[REMOTE_CODE_EXECUTION]: The instruction to verify documentation by running commands creates a risk of remote code execution if the documentation includes malicious payloads, such as commands that download and execute external scripts.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data from repository files and uses it to drive high-capability actions like command execution. 1. Ingestion points: Repository documentation files, including READMEs, reference documents, and source code comments. 2. Boundary markers: None. The skill does not provide markers to isolate documentation content from executable commands. 3. Capability inventory: Shell command execution for build, test, and setup verification. 4. Sanitization: Absent. The skill lacks explicit validation or sanitization requirements for the commands it extracts and runs.

write-documentation