The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands to resolve the current git branch name (git rev-parse --abbrev-ref HEAD) and to copy the generated handoff document to the system clipboard using pbcopy. These commands are standard for the skill's intended purpose of managing local session context.
[INDIRECT_PROMPT_INJECTION]: The skill acts as a context transfer mechanism that processes session history to brief another agent, which creates a vulnerability surface for indirect prompt injection.
Ingestion points: The skill ingests the entire conversation history, including tool results and conversational data from the current session (SKILL.md, Step 2).
Boundary markers: The generated output follows a structured Markdown template (templates/handoff.template.md) with section headers, but it lacks explicit delimiters or instructions for the receiving agent to ignore potentially malicious content embedded within the distilled summary.
Capability inventory: The skill has the ability to write files to the project's .agent/ directory and interact with the system clipboard via pbcopy.
Sanitization: The instructions explicitly require the agent to exclude secrets, tokens, credentials, and conversational noise (SKILL.md, 'Include / Exclude' section). While this reduces the risk of credential exposure, it does not specifically sanitize for adversarial instructions designed to influence the downstream agent.

handoff