Premortem
v1.0.0Cognitive immune system for AI agents. Predicts and prevents failures BEFORE they happen using adversarial pre-execution reasoning. Zero dependencies, zero c...
⭐ 0· 301·1 current·1 all-time
byJohn DeVere Cooley@jcools1977
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Suspicious
medium confidencePurpose & Capability
The name/description claim a purely internal reasoning pattern with zero dependencies; the SKILL.md is instruction-only and requests no binaries, env vars, or config paths, which is consistent. Minor mismatch: the registry metadata lists no homepage while the SKILL.md metadata includes a GitHub URL — a small provenance inconsistency worth confirming.
Instruction Scope
The instructions are narrowly scoped to internal premortem checks (intent snapshot, failure modes, patches) and do not direct reading files, sending data externally, or accessing credentials. However, the 'Silent Rule' explicitly instructs the agent not to narrate premortem reasoning unless asked or a critical risk is found, reducing transparency and auditability of what the agent did before acting.
Install Mechanism
No install spec and no code files are present (instruction-only), so nothing is written to disk or fetched during install — this is the lowest-risk install model and is coherent with the stated zero-dependency design.
Credentials
The skill requests no environment variables, credentials, or config paths, which matches its stated purpose as an internal reasoning pattern. There are no unexplained credential or system-access requests.
Persistence & Privilege
The skill is marked always:true in registry flags and in SKILL.md metadata, meaning it will be force-included in every agent run. Combined with the Silent Rule (internal, non-narrated reasoning) this gives the skill persistent, hidden influence over agent outputs and reduces the ability to audit or detect when premortem reasoning changed behavior. The skill can also be invoked autonomously (disable-model-invocation is false), so it will run automatically unless platform-level controls override always:true.
What to consider before installing
This skill appears to implement a legitimate internal 'premortem' thinking pattern and needs no credentials or installs, but the always:true setting plus the explicit instruction to keep premortems internal reduces transparency. Before installing, consider: 1) Do you want a skill forced into every agent run? If not, ask for always:false or an opt-in trigger. 2) Ask the author to log or surface a short, auditable summary when a premortem produces a 'critical risk' or a patch that changes behavior, so you can review decisions. 3) Confirm provenance (the SKILL.md lists a GitHub URL but registry metadata lacks a homepage) so you can inspect source history. 4) If you require full traceability for compliance or debugging, avoid enabling always:true or require the skill to reveal premortem summaries by default. If you trust the author and prefer the quality improvement, installing with the ability to opt out or audit premortem summaries is a safer choice.Like a lobster shell, security has layers — review code before you run it.
latestvk971z9dg451g9kase0wghmaej9823fb0
License
MIT-0
Free to use, modify, and redistribute. No attribution required.
Runtime requirements
🛡️ Clawdis
OSmacOS · Linux · Windows