The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted external data in the form of Spine Skeleton JSON and Atlas files. \n
Ingestion points: Files are read via discoverSpineAsset and parseAtlas in scripts/spine_assets.mjs and scripts/spine_atlas.mjs. \n
Boundary markers: The skill uses a structured manifest (schema v3) and strictly validates all input data against expected schemas. \n
Capability inventory: Filesystem operations are restricted to specific candidate directories using isWithin and realOrResolved checks. No network operations or arbitrary command execution capabilities are exposed. \n
Sanitization: Path normalization and validation are performed throughout the scripts to prevent directory traversal. SHA-256 hashing is used to ensure data integrity. \n- [COMMAND_EXECUTION]: The skill executes local Node.js scripts and uses the sharp library for image processing. No evidence of shell injection or arbitrary command execution was found. All inputs to sharp and filesystem commands are validated. \n- [EXTERNAL_DOWNLOADS]: The skill only uses the sharp library, which is a standard and trusted package for image processing. No remote scripts or binary downloads are performed at runtime. \n- [DATA_EXFILTRATION]: No network operations (e.g., curl, fetch) or attempts to access sensitive system files (e.g., .env, SSH keys) were identified. All data processing is local.

phaser4-spine-generative-reskin