The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external, potentially untrusted data sources including log files, network traces, codebase documentation (CONTEXT.md, ADRs), and test outputs. This creates a surface where malicious instructions embedded in those data sources could influence the agent's behavior.
Ingestion points: Diagnostic data from CONTEXT.md, ADRs, test results, CLI stdout/stderr, network traces, and browser logs.
Boundary markers: The skill does not specify the use of delimiters or specific instructions to the agent to disregard natural language commands found within the ingested diagnostic data.
Capability inventory: The skill utilizes broad system capabilities including shell command execution (curl, git), script execution, and browser automation.
Sanitization: The skill includes mandatory instructions to redact secrets and authentication tokens using the <REDACTED> placeholder before processing or displaying information.
[COMMAND_EXECUTION]: The skill requires the agent to execute a variety of debugging and diagnostic commands, such as git bisect, curl, and headless browser scripts (Playwright/Puppeteer), as well as custom bash scripts like the provided hitl-loop.template.sh for human-in-the-loop verification.

diagnosing-bugs