The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses local shell commands to gather context for the handoff document.
Evidence: scripts/handoff_self_check.py and scripts/handoff_template_generator.py utilize subprocess.run to execute git commands (e.g., git status, git log, git branch).
These operations are implemented using list-based arguments (avoiding shell=True) and are limited to read-only git status operations.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest and summarize conversation history, which creates a potential surface for indirect prompt injection.
Ingestion points: The skill processes the entire conversation history to generate the 5-section handoff document.
Boundary markers: While the AI is instructed to categorize and redact data via references/handoff_prompt.md, the scripts do not implement automated boundary markers for external data.
Capability inventory: The skill has the ability to read project configuration and write temporary markdown files.
Sanitization: The skill includes a dedicated scripts/redaction_linter.py script specifically designed to scan for and redact secrets, API keys, and PII before the document is finalized.

session-handoff