The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from Obsidian vault pages to synthesize instructions for newly generated agent skills.
Ingestion points: Processes content from pages within the directory specified by OBSIDIAN_VAULT_PATH to build the SKILL.md and references/ files.
Boundary markers: Lacks formal delimiters or instructions to ignore embedded prompts in the source data, though it advises the agent to "distil" rather than copy-paste.
Capability inventory: Includes file system write access to the review directory and the ability to execute local Python scripts.
Sanitization: There is no explicit sanitization or filtering mechanism for the data extracted from vault pages before it is interpolated into the generated skill's instructions.
[COMMAND_EXECUTION]: The skill executes external CLI tools and Python scripts to perform its tasks.
Evidence: Invokes qmd query for semantic page searching and executes several Python scripts: improve_description.py, package_skill.py, and quick_validate.py.
[DYNAMIC_EXECUTION]: The skill uses dynamic path resolution to locate and execute utility scripts at runtime.
Evidence: Determines the execution path for helper scripts by checking multiple possible locations within the $OBSIDIAN_WIKI_REPO structure (e.g., packaged install vs. source checkout) before execution.

vault-skill-factory