The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes Python's subprocess.run to interact with Git for retrieving branch information, commit history, and file status. These interactions are implemented using argument lists rather than raw shell strings, which follows best practices to prevent command injection vulnerabilities.\n- [INDIRECT_PROMPT_INJECTION]: The workflow involves reading markdown documents from the .claude/handoffs/ directory. This process presents a surface for indirect prompt injection if an agent were to process a handoff file containing malicious instructions. This risk is inherent to the skill's primary purpose of context restoration.\n
Ingestion points: Handoff files stored in .claude/handoffs/ are read during the RESUME workflow.\n
Boundary markers: The skill does not currently use specific markers to delimit handoff content from instructions.\n
Capability inventory: The agent can execute git commands and perform file operations via the skill's utility scripts.\n
Sanitization: The provided validation script focuses on completeness and secrets rather than filtering instructional content.\n- [SAFE]: The skill includes a dedicated validation script (scripts/validate_handoff.py) that proactively scans documents for sensitive information such as API keys, tokens, and private keys. This serves as a significant security control to prevent the accidental exposure of credentials within handoff documentation.

session-handoff