The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and act upon data from external documentation and HANDOFF.md files which may contain instructions or context from previous sessions.
Ingestion points: The agent is instructed to read docs/<feature>/HANDOFF.md, project orientation files (AGENTS.md, CONTRIBUTING.md), and authoritative documents (PLAN.final.md, EXECUTION.md).
Boundary markers: The skill explicitly warns the agent to "verify it before you trust it" and "trust the handoff exactly as much as it deserves," treating inputs as "claims, not facts."
Capability inventory: The skill uses standard git commands for repository inspection (status, log, diff, etc.) and invokes other local skills like execute-plan or execute-roadmap upon successful verification.
Sanitization: It implements a mandatory reconciliation process where every claim in the handoff is verified against the code and git history before being presented to the user for confirmation.
[COMMAND_EXECUTION]: All command execution is limited to standard, non-destructive git operations for repository inspection (git status, git log, git diff, etc.) used to verify the state of the work.
[SAFE]: The skill mandates a "Read, verify, propose, and ask" workflow, ensuring strict user oversight and confirmation before any modifications (such as merging or deleting the handoff file) are performed.
[SAFE]: No obfuscation, hardcoded credentials, or unauthorized network operations were detected. Data access is restricted to the project's documentation and source tree.

resume-handoff