The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes standard local development commands, specifically git status, git diff, and grep, to identify modified files and search for architectural patterns within the source code. These operations are restricted to the local repository and are essential for the skill's primary function of code verification.
[DATA_EXFILTRATION]: No network activity, data exfiltration patterns, or access to sensitive configuration files (such as credentials or SSH keys) were detected. All analysis is performed within the local environment.
[REMOTE_CODE_EXECUTION]: The skill does not download or execute code from external sources. All referenced documentation and scripts are located within the local filesystem.
[PROMPT_INJECTION]: The skill demonstrates an attack surface for indirect prompt injection as it ingests untrusted data from the local repository.
Ingestion points: The agent reads file names and source code content using git diff and grep -r in SKILL.md.
Boundary markers: No explicit delimiters or instructions to disregard embedded instructions are provided.
Capability inventory: The skill is limited to read-only repository analysis using git and grep.
Sanitization: No sanitization or validation of ingested file content is specified.

check-cross-layer