The Agent Skills Directory

[SAFE]: Evaluation of the skill's instructions and the provided script template indicates no malicious patterns, obfuscation, or unauthorized access to sensitive data. The skill's operations are well-aligned with its intended purpose for software debugging and performance regression analysis.
[COMMAND_EXECUTION]: The skill facilitates the use of various local command-line utilities and tools, such as curl for HTTP requests, git for bisection, and browser automation frameworks like Playwright or Puppeteer. These are standard practices for developers and agents performing diagnosis tasks.
[PROMPT_INJECTION]: The skill is exposed to indirect prompt injection due to its processing of external diagnostic data.
Ingestion points: Untrusted data enters the agent context from test outputs, server logs, and user-provided inputs as defined in SKILL.md and collected via scripts/hitl-loop.template.sh.
Boundary markers: Absent. There are no specific delimiters used to isolate untrusted data from the agent's core logic.
Capability inventory: The skill allows for network operations, file system access, and command execution, as described in SKILL.md.
Sanitization: Absent. The skill does not include steps to sanitize or filter input received from external files or user prompts.

diagnose