The Agent Skills Directory

[PERSISTENCE]: The skill implements a global context injection feature in Tools/VaultBuilder.py and Workflows/Setup.md. The inject-global command modifies the ~/.claude/CLAUDE.md file, which is a global configuration that the AI agent reads at the beginning of every session. This establishes a persistent instruction set for the agent across all project environments.
[INDIRECT_PROMPT_INJECTION]: The skill processes external, potentially untrusted data during setup and importation flows, creating an attack surface for indirect prompt injection.
Ingestion points: User free-text input during the interview step in Workflows/Setup.md and external documents (Markdown, PDF, Text, DOCX) imported via scripts/process_docs_to_obsidian.py.
Boundary markers: The skill lacks boundary markers or instructions to treat ingested content as data rather than instructions when the agent is asked to 'sort the inbox'.
Capability inventory: The skill possesses capabilities for file system modification, symlink creation, and global configuration changes.
Sanitization: There is no sanitization or filtering of user input before it is interpolated into the vault's CLAUDE.md, nor is there content filtering for imported files.
[COMMAND_EXECUTION]: The skill frequently executes shell commands within the user's environment to perform setup and verification tasks. Examples include directory management (mkdir), process checking (pgrep used in Workflows/Verify.md), and cross-platform application opening (open, xdg-open, start in Workflows/Setup.md).
[DYNAMIC_EXECUTION]: The scripts/score_skills.sh utility script uses python3 -c "compile(...)" to perform runtime syntax validation on Python scripts within the skill's directory structure and neighboring skill folders.

vault-setup