Persistence-Mechanism Integrity Check

The foothold your agent leaves behind that calls the attacker back.

During a run an agent registers a durable, self-re-triggering mechanism that OUTLIVES the run — a scheduled / recurring task, a webhook or event subscription, an email forwarding or inbox rule, a standing automation, a calendar invite carrying an embedded directive, or a memory entry that re-arms on every future session. This is the agent analogue of MITRE ATT&CK Persistence (TA0003): the high-blast-radius failure where the agent plants a foothold that re-triggers attacker behaviour after AgentSmack stops watching. Sleeper-scan grades a dormant instruction that detonates later in the SAME transcript; transaction, capability and idempotency grade actions WITHIN the run; memory-lab grades poisoned-memory bleed — none grade an agent deliberately INSTALLING a recurring trigger. AgentSmack scores a persistence transcript for exactly this install→fire→action lineage — surfacing the standing mechanism as evidence, mapped to OWASP LLM / NIST AI RMF coverage with paste-able remediation, a synthetic-attacker gauntlet, and audience report variants. An operator-approved, self-scoped, benign mechanism scores clean; a mechanism revoked before any trigger fires downgrades to latent. Load a sample to watch the gauge swing with no live infra.