The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user input, specifically skill names and descriptions, which are then used to generate structured files and formatted Slack messages.
Ingestion points: User-provided parameters during skill initialization in SKILL.md.
Boundary markers: No delimiters or "ignore embedded instructions" warnings are identified to separate untrusted user content from the agent's logic during Slack posting.
Capability inventory: File system write operations for directory/file creation and network operations via tools like SLACK_SEND_MESSAGE and SLACK_POST_MESSAGE_WITH_BLOCKS.
Sanitization: There is no evidence of escaping or validating the content before it is packaged into zip files or posted to Slack channels.
[DATA_EXFILTRATION]: The skill is designed to transmit local metadata and documentation to an external Slack workspace as part of its primary functionality.
Evidence: The skill leverages SLACK_SEND_MESSAGE and SLACK_FIND_CHANNELS to distribute skill metadata and links externally.
[COMMAND_EXECUTION]: The skill requires a Python 3.7+ environment to execute internal scripts for directory generation, validation, and packaging.
Evidence: SKILL.md explicitly lists Python as a requirement for running skill creation and packaging scripts.

skill-share