The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes session logs which may contain data from previous untrusted sources processed in past interactions (e.g., contents of web pages or external code).
Ingestion points: Reads .jsonl files from ~/.claude/projects/ and ~/.codex/sessions/, as well as thread activity via t3_thread_read in Step 2 and 3.
Boundary markers: None identified; subagents process the log content to find 'friction' without explicit instructions to ignore embedded commands.
Capability inventory: The skill uses git worktree list, reads repository configuration files (such as package.json), and has read access to the project files.
Sanitization: No sanitization or filtering of the log content is described before it is passed to subagents for analysis.
[DATA_EXPOSURE]: The skill is designed to access and read session interaction history, which contains a record of developer activity and project context.
Evidence: Accesses ~/.claude/projects/<dir>/*.jsonl and ~/.codex/sessions/YYYY/MM/DD/*.jsonl to extract session metadata and activity logs.

retro