The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on the execution of a locally installed CLI binary located at .slipbox/bin/slipbox. Evidence in SKILL.md identifies this CLI as the owner of atomic filesystem operations and the vault's runtime contract.
[COMMAND_EXECUTION]: The skill performs Git operations and triggers repository hooks during the finalization process. Evidence in references/git-finalization.md describes invoking slipbox work commit and notes that "normal Git hooks run."
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and store external text content verbatim, which constitutes a potential surface for indirect prompt injection.
Ingestion points: reader-approved propositions as described in references/evergreen-candidates.md and Resource analysis in references/work-lifecycle.md.
Boundary markers: The instructions do not specify delimiters or safety warnings to prevent the agent from processing embedded instructions within the verbatim text.
Capability inventory: The skill can execute a local CLI tool, modify the local filesystem, and commit changes to a Git repository.
Sanitization: The skill specifically requires recording propositions "exactly as supplied" without paraphrasing or inference, meaning no sanitization or escaping is performed on external data.

using-slipbox