The Agent Skills Directory

[DATA_EXPOSURE]: The skill accesses session logs located at ~/.claude/projects/*/<sessionId>.jsonl. These transcripts contain full conversation histories, tool outputs, and potentially sensitive code or environment information.
The skill includes specific instructions to "Never leak secrets" and "No secrets / tokens / PII" to mitigate risks when creating the summary.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data in the form of previous session transcripts.
Ingestion points: Reads .jsonl files from the Claude projects directory (SKILL.md).
Boundary markers: Instructs the agent to filter out tool_result blocks and focus on narrative content, but lacks formal delimiters for the raw log content during summarization.
Capability inventory: File read (logs), File write (markdown records), and Bash execution via the fork's toolset.
Sanitization: Instructions mandate summarization and exclusion of secrets, but the agent remains susceptible to malicious instructions embedded within the processed history logs.
[EXTERNAL_DOWNLOADS]: The skill references official documentation from Claude Code to verify platform-specific environment variables and file path conventions.
Fetches configuration and syntax details from the official code.claude.com documentation site.

session-to-meeting-log