The Agent Skills Directory

[DATA_EXPOSURE]: The skill accesses sensitive local session transcripts located in ~/.claude/projects/. This data contains the history of user and assistant interactions, which is required for the skill's audit functionality but involves private information.
[COMMAND_EXECUTION]: The instructions direct the agent to use a custom CLI tool called orbit-audit. This tool is used to perform codebase validation and compare project plans against evidence found in transcripts.
[INDIRECT_PROMPT_INJECTION]: The skill parses untrusted data from local session history files, creating a surface for potential indirect prompt injection attacks.
Ingestion points: ~/.claude/projects/**/*.jsonl files.
Boundary markers: Absent. The instructions do not specify any delimiters or warnings to prevent the agent from obeying instructions found within the transcripts.
Capability inventory: The skill is capable of executing shell commands via orbit-audit and reading arbitrary files in the repository for validation.
Sanitization: Absent. There is no evidence of filtering or escaping transcript content before it is processed by the agent.

gap-analysis