The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill installs a CLI tool from a GitHub repository owned by the author. Evidence: uv tool install git+https://github.com/cearley/claude-session-index in SKILL.md.
[COMMAND_EXECUTION]: The skill utilizes shell commands to interact with the installed sessions CLI to query a local database. Evidence: Examples provided include sessions context, sessions find, and sessions analytics.
[DATA_EXFILTRATION]: The skill indexes and accesses local directories containing sensitive Claude Code session history and conversation metadata. Evidence: Accesses ~/.claude/projects/, ~/.claude/session-topics/, and ~/.session-index/.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it reads and summarizes historical conversation logs which may contain untrusted data or instructions from past sessions.
Ingestion points: Historical data is ingested via sessions context <session_id>.
Boundary markers: Absent; no specific instructions are provided to the agent to distinguish historical text from system instructions.
Capability inventory: The skill can execute shell commands (sessions) and spawn sub-agent synthesis tasks.
Sanitization: Absent; data from logs is processed directly without filtering or escaping.

session-index