The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes several shell utilities including curl, cp, find, and identify. It also runs a local validation script scripts/canvas_validate.py on the resulting JSON files.
[REMOTE_CODE_EXECUTION]: The skill uses a Python one-liner python3 -c "from PIL import Image; img=Image.open('[path]'); ..." to determine image dimensions. The [path] variable is interpolated directly into the Python command string. If [path] contains unsanitized characters like single quotes, an attacker could execute arbitrary Python code in the host environment.
[EXTERNAL_DOWNLOADS]: The add image function uses curl -sL to download files from arbitrary URLs provided by the user or external sources. This lack of URL validation could lead to the download of malicious content or Server-Side Request Forgery (SSRF).
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes untrusted external data (URLs and remote filenames) and incorporates them into execution paths.
Ingestion points: External URLs and local file paths in SKILL.md.
Boundary markers: None identified.
Capability inventory: Shell command execution (curl, find), Python execution, and file system write access in SKILL.md.
Sanitization: No evidence of escaping or validation for variables used in command interpolation.

canvas-populate