The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes lychee and git via subprocess.run() to perform link checking and repository path resolution. It also optionally invokes a vendor-provided script article_status.py to record run results. These are standard operations for a development-focused link-checking utility.
[REMOTE_CODE_EXECUTION]: While the script recommends installing the lychee tool via brew install lychee if missing, it does not automatically download or execute arbitrary remote code. The execution is limited to local binaries already in the user's PATH.
[DATA_EXFILTRATION]: The network operations are initiated by the external lychee tool (a standard link checker) to validate URL reachability. The skill itself does not transmit sensitive local data to external servers.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external URLs and server responses (404s, timeouts). It implements a structured triage process that maps these inputs to a fixed set of verdicts (broken, unverified, known false positive) before displaying them to the agent, which limits the risk of raw data influencing the agent's logic.

write-check-links