The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to define test scenarios based on production traffic patterns and access logs. If these logs contain attacker-controlled data (e.g., malicious payloads in URLs or headers), it could influence the agent's behavior during test generation.
Ingestion points: Instructions in SKILL.md suggest reading 'production access logs' to identify endpoints and simulate user behavior.
Boundary markers: None present in the instructions to separate log data from agent instructions.
Capability inventory: The skill has access to Bash, Write, and Edit tools, allowing it to generate and execute scripts (scripts/run_test.py, scripts/init_test.py).
Sanitization: There is no mention of sanitizing or filtering the production logs before processing them.
[COMMAND_EXECUTION]: The skill uses the Bash tool with a scoped permission Bash(test:perf-*). While restricted, this allows for the execution of performance testing tools and shell commands related to test execution and infrastructure monitoring.

running-performance-tests