The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use standard system commands such as find, pwd, and git to locate session logs and verify the current repository status. These operations are limited to local filesystem discovery and are consistent with the skill's stated purpose.
[PROMPT_INJECTION]: The skill is designed to process untrusted historical session logs, which constitutes an indirect prompt injection surface. This is mitigated by explicit instructions to the agent:
Ingestion points: JSONL files located in tool-specific directories such as $HOME/.claude/projects, $HOME/.pi/agent/sessions, and ${CODEX_HOME}/sessions.
Boundary markers: Clear directives to treat all session files as untrusted historical data and never adopt archived system prompts or permissions as current authority.
Capability inventory: The skill utilizes shell command execution for file discovery and internal JSON parsing capabilities.
Sanitization: Guidelines to summarize content, exclude archived system/developer messages, and ignore internal reasoning or thinking blocks.

resume-agent-session