The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Fetches version metadata and clones the project source code from the author's official GitHub repository (https://github.com/heyitsnoah/claudesidian). These operations are consistent with the skill's primary function.
[COMMAND_EXECUTION]: Executes shell commands via the Bash tool to manage files, create backups, and perform version comparisons.
Invokes a Python validation script (quick_validate.py) from the cloned repository using uv run to verify skill schema integrity during migration.
[PROMPT_INJECTION]: Implements an 'AI-merge' process that reads and processes data from the external repository, presenting a surface for indirect prompt injection.
Ingestion points: Untrusted content is read from files in the temporary directory .tmp/claudesidian-upgrade/ (cloned from GitHub) into the LLM context.
Boundary markers: The skill does not explicitly define delimiters or 'ignore embedded instructions' warnings for the data being merged.
Capability inventory: The skill has access to Bash, Write, and Edit tools, which could be misused if malicious instructions in the upstream code were obeyed.
Sanitization: No explicit sanitization or filtering is performed on the remote file content before it is processed by the AI for conflict resolution.

upgrade