The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses local shell commands, primarily git and ls, to identify the repository root, determine the current branch, and fetch commit history. These operations are essential for its primary purpose of anchoring documentation to specific code changes.
[INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection as it ingests untrusted data from Git commit logs and verbatim operator quotes, which are then written into Markdown files. If these files are later processed by an AI agent, the embedded content could influence behavior.
Ingestion points: Reads output from git log, git status, and verbatim input provided by the operator during the session (as described in SKILL.md).
Boundary markers: Uses structured Markdown headers (e.g., ## Решения, ## Коррекции) and specific line prefixes (e.g., DECIDED [operator]) to delimit content.
Capability inventory: The skill has file system write capabilities within the .trail/entries/ directory and executes read-only Git commands.
Sanitization: The skill instructions explicitly demand verbatim quoting of operator input without translation or paraphrasing, which means no sanitization is performed on user-supplied strings before they are written to the documentation files.

trail-record