The Agent Skills Directory

[SAFE]: The skill is a standard developer tool designed for troubleshooting E2E tests. No malicious patterns or security risks were identified.
[EXTERNAL_DOWNLOADS]: The skill includes functionality to download test artifacts from GitHub Actions.
Evidence: scripts/fetch-gh-run.mjs uses the gh CLI (gh run download) to fetch ZIP files from official GitHub repositories.
Context: This is a core feature of the skill and targets a well-known service (GitHub), adhering to expected behavior.
[COMMAND_EXECUTION]: The skill executes shell commands to process trace files.
Evidence: scripts/fetch-gh-run.mjs and scripts/trace-extract.mjs use node:child_process to call gh, unzip, and python3.
Mitigation: These scripts use execFileSync with argument arrays rather than shell strings, effectively preventing command injection vulnerabilities.
[PROMPT_INJECTION]: The skill analyzes external data (test logs and error messages) which could theoretically contain instructions.
Ingestion points: The agent reads content from trace.trace and trace.network JSONL files, including console logs and error stack traces.
Boundary markers: While explicit delimiters for the trace data are not defined, the skill focuses on technical measurement and root-cause analysis rather than blindly following instructions found in the data.
Capability inventory: File system access and shell execution are limited to the provided scripts.
Sanitization: Content is parsed as structured JSONL before analysis.
Note: This represents a standard attack surface for any analysis tool; no actual injection patterns were detected in the skill instructions.

playwright-trace-analyzer