The Agent Skills Directory

[SAFE]: The skill is a template-driven instructional document focused on memory management. It does not contain executable code, remote downloads, or credential exposure.
[CREDENTIALS_UNSAFE]: The skill explicitly mandates that session resumption must NEVER load raw API keys or secrets from text files, requiring re-authentication with a proper Secret Manager (e.g., HashiCorp Vault). This follows industry best practices for secret management.
[INDIRECT_PROMPT_INJECTION]: The skill identifies the risk of 'Amnesia Exploitation' where a malicious actor might supply a backdoored snapshot. It mitigates this by requiring snapshots to be chronologically continuous and cryptographically signed before loading.
[PRIVILEGE_ESCALATION]: No privilege escalation patterns were detected. The skill operates within the agent's defined file system scope (agents/memory/).
[REMOTE_CODE_EXECUTION]: No remote code execution or external network operations were detected. All file access is local to the project memory directory.

session-resume