The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell utilities including tail, jq, find, and sort to extract session metadata and locate scripts within the ~/.claude configuration directory.
[DYNAMIC_EXECUTION]: The skill performs runtime discovery of a local shell script (half-clone-conversation.sh) using the find command and subsequently executes it. This execution pattern relies on the script being present in the user's home directory and assumes the file's integrity.
[INDIRECT_PROMPT_INJECTION]: The skill interacts with conversation history files, which serve as a source of untrusted data from previous session turns.
Ingestion points: The skill reads metadata from ~/.claude/history.jsonl using tail and jq (SKILL.md, Step 1).
Boundary markers: None identified; the instructions do not include specific delimiters or warnings to ignore instructions that might be embedded within the historical data.
Capability inventory: The skill has access to shell execution (bash, tail, find, jq) and the ability to execute identified local scripts.
Sanitization: The skill extracts specific JSON fields using jq but does not explicitly sanitize or validate the content of the sessionId or project fields before they are passed as arguments to other commands.

half-clone