The Agent Skills Directory

[SAFE]: The skill is designed with a strong emphasis on data integrity and verification. It provides comprehensive validation scripts to ensure consistency between human-readable handoff files and structured JSON context packs.
[COMMAND_EXECUTION]: The skill includes several helper scripts (build_handoff_index.py, validate_context_pack.py, validate_handoff_md.py, validate_handoff_parity.py) and a test runner (smoke_test.sh). These scripts use standard library modules like argparse, pathlib, and hashlib. They operate strictly on local files and do not invoke external processes or shell commands in an unsafe manner.
[DATA_EXPOSURE]: The indexing script reads local workspace artifacts such as plan.md and result.md to calculate hashes and summaries. This is the primary function of the skill and does not involve data exfiltration, as no network-capable functions are used.
[INDIRECT_PROMPT_INJECTION]: The skill processes external Markdown files to generate handoff summaries, which represents a potential surface for indirect prompt injection.
Ingestion points: The skill reads plan.md, progress.md, task.md, and result.md from the current workspace.
Boundary markers: Not explicitly defined in the prompt instructions, though the skill uses structured JSON and Markdown tables for output.
Capability inventory: The agent's actions are restricted to reading files and generating documentation; no network or privilege-related tools are involved.
Sanitization: The skill implements strict structural validation through its parity and schema scripts, which helps prevent malformed data from affecting the handoff process.

workspace-handoff-compiler