The Agent Skills Directory

[DATA_EXFILTRATION]: The skill is specifically designed to locate, read, and analyze sensitive session data including main logs (~~/.claude/projects/.jsonl), debug logs (~/.claude/debug/.txt), and subagent transcripts (~~/.claude/projects//agent-.jsonl). These files contain a comprehensive record of user interactions, internal reasoning, tool calls, and potentially sensitive data or secrets that were processed or output during previous sessions.
[COMMAND_EXECUTION]: The skill relies on multiple shell scripts to automate file discovery and data extraction:
scripts/find-session-files.sh searches the ~/.claude directory for logs and transcripts.
scripts/extract-subagent-calls.sh uses grep to parse execution events from debug logs.
scripts/extract-hook-events.sh extracts hook results and model responses from debug logs.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted data from previous session logs to verify behavior. This creates a surface for indirect prompt injection if a previous session contained malicious content designed to influence the analyzer's output.
Ingestion points: ~/.claude/debug/{sessionId}.txt and ~/.claude/projects/.../{sessionId}.jsonl are read by analysis scripts.
Boundary markers: No explicit boundary markers or instructions to ignore embedded commands are present in the scripts or instructions.
Capability inventory: The skill uses Bash to run analysis scripts and has the capability to write report files to the filesystem.
Sanitization: No sanitization or validation of the log content is performed before parsing or report generation.

session-analyzer