The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/scan-secrets.mjs invokes execFileSync to run git ls-files and git check-ignore. These commands are used to verify whether sensitive files are tracked by git or ignored by the project's configuration during the security scan.
[DYNAMIC_EXECUTION]: In scripts/lib/browser.mjs, the skill uses dynamic import() to load the playwright or playwright-core library from a computed path. Additionally, it searches for browser executables across various system and user-specific locations, including user cache directories (e.g., ~/Library/Caches/ms-playwright), which is a form of dynamic environment discovery.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves ingesting and processing untrusted data from the site being audited, which presents a surface for indirect prompt injection attacks.
Ingestion points: scripts/lib/pages.mjs fetches site content through web crawling and headless browser rendering; scripts/scan-secrets.mjs recursively reads local source code and configuration files.
Boundary markers: The instructions do not provide the agent with specific delimiters or "ignore instructions" warnings to apply when processing the external data it collects.
Capability inventory: The skill possesses extensive capabilities, including reading local files, making external network requests, executing git commands, and controlling a headless browser.
Sanitization: The collected data is parsed and reported to the agent without explicit sanitization or filtering to prevent the execution of instructions embedded in the audited content.
[CREDENTIALS_UNSAFE]: scripts/scan-secrets.mjs contains comprehensive regex patterns for detecting hardcoded credentials, such as AWS access keys, Stripe secret keys, and GitHub tokens. It also accesses environment files like .env. Although this behavior is intended for auditing and includes redaction logic to prevent full secrets from being reported, the ability to read these credentials represents a sensitive operation.

web-security