The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to perform systematic crawls of 500-1000 external website examples, which exposes the agent to untrusted data that could contain malicious instructions.
Ingestion points: External data is ingested from untrusted third-party websites using WebFetch and WebSearch as described in the Core Process section of SKILL.md.
Boundary markers: The skill lacks explicit instructions for using delimiters or boundary markers to isolate external content, increasing the risk that the agent may follow instructions embedded within the scraped design examples.
Capability inventory: The skill possesses file write capabilities (Write tool for checkpoints) and network access (WebFetch/WebSearch), which could be misused if the agent is influenced by injected content.
Sanitization: There are no defined procedures for sanitizing or filtering instructions from the HTML and visual data extracted during the crawl.
[COMMAND_EXECUTION]: The skill package includes a shell script (scripts/validate_archive.sh) intended for local execution to verify output integrity.
Evidence: The script uses the jq utility to perform structural and logic checks on the generated design database. It does not perform network operations or require elevated privileges, serving as a static validation utility.

design-archivist