The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs automated file system operations to create and manage new skill directories. Specifically, Phase 5 of the SKILL.md workflow instructs the agent to create a specific directory structure under skills/[skill-name]/, generate several files (SKILL.md, openai.yaml, style-guide.md), and copy the generate_image.py script. It also references the execution of a quick_validate.py script to verify the generated skill structure.
[DATA_EXFILTRATION]: The script scripts/generate_image.py performs network requests to https://generativelanguage.googleapis.com to generate images. This targets a well-known service from Google and is used for the skill's primary intended purpose of image generation. Additionally, the file references/skill-template.md contains a hardcoded absolute file path: /Users/ugreen/Documents/obsidian/09image/. This exposes details of the author's local file system structure and may lead to execution failures or unexpected behavior on systems where this path does not exist.
[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection. It collects user-provided strings through interactive questions and interpolates them directly into a generated SKILL.md file using the references/skill-template.md. There is no evidence of sanitization or validation of this user-provided content before it is written to the new skill's instructions. * Ingestion points: User responses to questions Q1 through Q10 in the SKILL.md workflow. * Boundary markers: None are present to delimit user input or instruct the agent to ignore embedded commands within the responses. * Capability inventory: The skill possesses the capability to write files to the local disk and execute scripts (generate_image.py, quick_validate.py). * Sanitization: No sanitization, escaping, or validation of user-provided strings is implemented during the skill generation phase.

space-image-skill-builder