The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill reads external files, including source code (.html, .css, .blade.php) and test results (test-results.json), which serves as an ingestion point for untrusted data.\n
Ingestion points: The agent uses grep on source files and cat on public/qa-screenshots/test-results.json to gather data for its report.\n
Boundary markers: No delimiters or specific instructions are provided to distinguish between the agent's instructions and the content of the files being analyzed.\n
Capability inventory: The skill allows the agent to execute shell commands, including running local scripts and file system operations.\n
Sanitization: There is no evidence of sanitization or validation performed on the data ingested from the file system before it is processed by the agent.\n- [COMMAND_EXECUTION]: The skill relies on the execution of shell commands and local scripts to perform its primary function.\n
Evidence: SKILL.md explicitly instructs the agent to run ./qa-playwright-capture.sh, ls -la, grep, and cat as part of its mandatory process.\n
Context: While these commands are aligned with the skill's stated purpose of quality assurance, they represent a capability that could be abused if the agent is influenced by malicious input.

testing-evidence-collector