The Agent Skills Directory

[COMMAND_EXECUTION]: The flag_debt_scanner.py script executes git log via subprocess.run to analyze the commit history and determine the age of feature flags. The execution is performed securely using an argument list instead of a shell string, and flag identifiers are validated against a restrictive regex.
[INDIRECT_PROMPT_INJECTION]: The skill scans repository source code files to identify feature flag identifiers, which constitutes a surface for indirect prompt injection if malicious content were embedded in code.
Ingestion points: The flag_debt_scanner.py and kill_switch_audit.py scripts walk the repository directory tree and read the contents of files with common code extensions (e.g., .py, .js, .ts).
Boundary markers: The scripts process file contents to extract identifiers but do not explicitly use boundary markers or provide instructions to the agent to ignore potential instructions within those files.
Capability inventory: The skill scripts are restricted to local file reading and executing git for history analysis.
Sanitization: Extracted identifiers are sanitized using a strict regular expression ([\w.\-:]+) that limits identifiers to alphanumeric characters, dots, hyphens, and colons before they are used in any command execution.

feature-flags-architect