The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its core functionality of syncing instructions (CLAUDE.md) from remote or external sources into multiple local project environments.
Ingestion points: Fetches content from a default remote URL (https://raw.githubusercontent.com/cleaniquecoders/kickoff/refs/heads/main/stubs/CLAUDE.md) or user-specified local/remote paths via the --source flag (SKILL.md, references/registry-schema.md).
Boundary markers: The merge algorithm identifies sections by header boundaries but does not wrap the imported content in safety delimiters or warnings to prevent the agent from following malicious instructions embedded in the source (references/merge-algorithm.md).
Capability inventory: The skill possesses the ability to write to the filesystem (overwriting CLAUDE.md), execute git commands (add, commit), and perform broad filesystem searches (find) (SKILL.md).
Sanitization: The skill lacks sanitization or validation logic for the content being merged; it treats the source content as a trusted set of instructions without verifying for adversarial prompts (references/merge-algorithm.md).
[EXTERNAL_DOWNLOADS]: The skill automatically fetches convention stubs from an external GitHub repository (cleaniquecoders/kickoff) to use as the 'source of truth' for local projects (SKILL.md).
[COMMAND_EXECUTION]: The skill invokes several CLI tools to perform its tasks, including find for project discovery, git for committing changes, and the GitHub CLI (gh) for remote repository scanning and metadata extraction (SKILL.md, references/detection-markers.md).

project-sync