The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands including find, cat, and node -e to locate and extract data from local files.
Evidence: Phase 1 and Phase 3 utilize multiple shell one-liners to process mochawesome.json and results.xml files.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it ingests untrusted data from test reports.
Ingestion points: The skill reads test titles, error messages, and stack traces from mochawesome.json and results.xml in SKILL.md.
Boundary markers: There are no delimiters or "ignore embedded instructions" warnings used when interpolating report data into the agent's context.
Capability inventory: The agent has access to file reading, shell execution (find, cat, node), and browser tools for inspecting screenshots/videos as described in Phase 1 and 3.
Sanitization: No sanitization, escaping, or validation of the report content is performed before the data is processed by the AI.

cypress-debugger