The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources during its audit operations.\n
Ingestion points: The skill ingests content from user-provided URLs via a browser tool, reads source code files from the local directory, and processes image data from screenshots as outlined in SKILL.md.\n
Boundary markers: There are no explicit instructions or delimiters defined to prevent the agent from following potential instructions embedded in the audited content.\n
Capability inventory: The skill possesses the capability to execute shell commands, perform file system reads, and conduct network operations via an agent-controlled browser.\n
Sanitization: No evidence of input sanitization or filtering was found for the data retrieved from external URLs or source files.\n- [COMMAND_EXECUTION]: The skill utilizes shell commands for environment detection and project navigation.\n
Evidence: In SKILL.md, the skill uses git diff to identify changes in UI-related files and the find command to discover the structure of the input codebase.\n- [DYNAMIC_EXECUTION]: The skill executes external scripts from the local file system to load configurations.\n
Evidence: The skill uses the source command in SKILL.md to execute read-config.sh from a relative path within the stelow-workflow-orchestrator project structure, which is a resource owned by the same author.\n- [EXTERNAL_DOWNLOADS]: The skill performs network operations to access live sites for auditing.\n
Evidence: The "Live Site Audit" mode in SKILL.md uses a browser tool to fetch and interact with content from user-specified URLs.

stelow-workflow-ux-critique