Conatus
PassAudited by ClawScan on May 1, 2026.
Overview
Conatus is an instruction-only self-analysis skill with disclosed optional memory logging and scheduled reflections, with no hidden code, credentials, or destructive behavior shown.
This skill appears safe to install as an instruction-only self-analysis aid. Before enabling the optional heartbeat or daily reflection workflows, decide where its memory logs should live, how long they should be kept, and whether they should include any sensitive task or user information.
Findings (2)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
The agent may keep ongoing self-assessment logs that could influence later behavior or expose information included in memory files.
The skill encourages storing recurring self-analysis output in persistent agent memory, which is aligned with the stated purpose but can retain inaccurate or sensitive context if not scoped.
Log results to memory/conatus-log.md
Keep Conatus logs in a clearly scoped folder, review or delete them periodically, and avoid storing sensitive user or task details unless needed.
If configured, the agent may spend recurring attention on self-reports and alerts even when not directly asked in the moment.
The skill suggests recurring self-analysis through heartbeat integration. This is disclosed and purpose-aligned, but it is still persistent agent behavior if the user enables it.
Run conatus self-analysis every 4th heartbeat
Only add the heartbeat or daily schedule if desired, set a clear frequency and stop condition, and make sure the generated reports are useful enough to justify the recurring activity.