The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local shell commands including git, jq, and gh (GitHub CLI) to extract metrics such as commit counts, lines changed, and pull request statuses. These operations are essential for the skill's primary purpose of generating engineering reports and do not involve remote command execution or unsafe piping from external sources.- [INDIRECT_PROMPT_INJECTION]: The skill processes data from Git commit messages which could theoretically contain malicious instructions.
Ingestion points: Commit messages are ingested via git log and git diff commands as specified in SKILL.md and reference/metrics-guide.md.
Boundary markers: The skill does not implement specific delimiters or instructions to ignore embedded content when processing Git logs.
Capability inventory: The skill has the capability to write JSONL data to ~/.claude/portfolio/ and generate HTML artifacts.
Sanitization: Commit data is primarily sanitized by being passed through counting utilities like wc -l or filtered for specific conventional commit prefixes (feat:, fix:) using grep, which effectively mitigates the risk of instructions being interpreted by the agent.

portfolio-artifact-skill