The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides analysis and packaging tools that ingest and process SKILL.md files from external directory paths provided at runtime.
Ingestion points: The scripts analyze_skill.py, upgrade_skill.py, and package_skill.py read and parse the content of SKILL.md files from local filesystem paths.
Boundary markers: There are no explicit boundary markers or "ignore instructions" warnings implemented when the scripts present the analyzed content back to the agent.
Capability inventory: The skill has the capability to read files, create directories (init_skill.py), and write zip archives to the filesystem (package_skill.py).
Sanitization: While the scripts use yaml.safe_load() to securely parse frontmatter, the markdown body is processed using regular expressions and the results are presented to the agent, which could be influenced by malicious instructions embedded in the analyzed skill files.
[PRIVILEGE_ESCALATION]: The scripts/init_skill.py script programmatically modifies file permissions.
Evidence: The script uses example_script.chmod(0o755) to grant execution permissions to a dynamically created example script during the skill initialization process.
[COMMAND_EXECUTION]: The skill instructions in SKILL.md require the agent to execute several local Python scripts to perform its core functions.
Evidence: Instructions include commands such as python scripts/analyze_skill.py <path/to/skill> and scripts/init_skill.py <skill-name> --path <output-directory>. This is consistent with the skill's purpose as a developer toolset.

skill-creator-plus