The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool to perform shell operations for artifact detection and file discovery. As specified in the 'Non-Negotiable Contract' and 'Step 1' sections of SKILL.md, it executes git diff with grep to identify architectural changes and uses find to locate design documents within the project directories.
[REMOTE_CODE_EXECUTION]: In SKILL.md (Step 3) and the reference file 2026-07-17_execute-logic-bearing-specs.md, the agent is instructed to perform an 'Empirical pass' where it 'drives the real implementation' or executes test harnesses for logic described in the documents under review. This involves dynamic execution of project code at runtime to verify architectural hypotheses.
[EXTERNAL_DOWNLOADS]: The skill utilizes WebFetch (SKILL.md, Step 1) to retrieve architecture documents or specifications from external URLs provided by the user. The generated HTML playground also involves loading resources from external CDNs.
[DATA_EXFILTRATION]: The skill performs network operations using WebFetch to fetch content from non-whitelisted domains during its document retrieval phase. While this is part of its core functionality, it represents a network communication capability.
[PROMPT_INJECTION]: The skill processes untrusted data from external documents and URLs. The 'Empirical pass' capability creates an attack surface for indirect prompt injection, where a malicious document could influence the agent's execution of commands or logic verification. The skill does not explicitly define sanitization or boundary markers for data processed during this empirical execution phase.

wk-arch-review