The Agent Skills Directory

DATA_EXFILTRATION (MEDIUM): In references/codebase-understanding.md, the skill explicitly directs the agent to 'Analyze dotenv files and configuration'. Dotenv files frequently contain unencrypted secrets, API keys, and database credentials, posing a significant risk of sensitive data exposure.
COMMAND_EXECUTION (MEDIUM): The skill relies on executing several external commands and local scripts:
It executes node .claude/scripts/set-active-plan.cjs to manage session state.
It uses repomix --remote <github-repo-url> in references/research-phase.md to download and process remote repository content.
It uses the gh CLI to read logs, pull requests, and issues.
INDIRECT_PROMPT_INJECTION (LOW): The skill has a high surface area for indirect injection as it processes untrusted content from remote repositories and GitHub issues.
Ingestion points: Remote GitHub URLs via repomix, GitHub issues/PRs via gh, and external documentation via docs-seeker.
Boundary markers: None explicitly defined in the provided instructions to delimit untrusted data.
Capability inventory: File system access (reading/writing plans), environment variable modification, and shell command execution.
Sanitization: No sanitization or validation of the ingested data is specified before it is used to generate implementation plans.

planning