The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/audit-brand.mjs uses spawnSync to execute the external magick command (ImageMagick) to retrieve image dimensions.
Evidence: spawnSync("magick", ["identify", "-format", "%w %h", filePath], ...) in scripts/audit-brand.mjs.
[DATA_EXFILTRATION]: The script reads and processes files based on paths provided in a configuration file (.brandfy/config.yaml). A malicious configuration could potentially direct the script to access sensitive local files, exposing their existence, hashes, or metadata in the generated audit.md report.
Evidence: The main function in scripts/audit-brand.mjs resolves paths for brand_directory, brand_document, and chromatic_guide from the config file and performs readFile and sha256 operations on them.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted project data (manuals, CSS, SVG, JSON) which could contain instructions intended to influence the agent's behavior during the audit process.
Ingestion points: The script scripts/audit-brand.mjs reads various project files including .brandfy/config.yaml, BRAND.md, CHROMATIC.md, and assets in the brand directory.
Boundary markers: The instructions in SKILL.md and references/audit-rubric.md do not provide explicit delimiters or warnings to ignore instructions embedded within the files being audited.
Capability inventory: The skill possesses file read/write capabilities (readFile, writeFile) and command execution capabilities (spawnSync).
Sanitization: There is no evidence of sanitization or filtering of the content read from external files before it is processed or included in the report.

brandfy-auditoria