The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests content from multiple untrusted sources which could be used to influence the agent's decision-making or suggested tasks.
Ingestion points: Reads contents from project-specific files like learnings.md, HANDOVER.md, ROADMAP.md, and FUTURE_FEATURES.md. It also fetches external data using the GitHub CLI (gh).
Boundary markers: No explicit delimiters or instructions are provided to the agent to treat the contents of these files as untrusted data.
Capability inventory: The skill has broad execution capabilities including npm, python3, git, and gh.
Sanitization: The skill does not perform validation or sanitization on the external content it reads before processing it.
[COMMAND_EXECUTION]: The orientation routine relies on the automated execution of several shell commands in the local environment.
Evidence: Regularly executes git status, git branch, npm test, npm run build, npm run lint, and gh issue list to gather project metadata and health status.
[DATA_EXFILTRATION]: The skill accesses the ~/.claude/ directory, which is a hidden folder containing agent-maintained project memory and global configurations.
Evidence: Searches for and reads MEMORY.md and architecture topic files within ~/.claude/projects/<escaped-project-path>/memory/.

session-start