Imprint
v1.0.1Adaptive operator modeling for AI agents. Your agent learns who you are by watching — not by being told. Builds a predictive model of your preferences, patte...
⭐ 0· 81·0 current·0 all-time
byShadow Rose@theshadowrose
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
medium confidencePurpose & Capability
The name/description (adaptive operator modeling) aligns with the instructions: the skill describes observing operator behavior, building a local operator-model.json, and using it to anticipate actions. It requests no external credentials, binaries, or installs, which is proportionate to the stated purpose.
Instruction Scope
SKILL.md directs the agent to observe user interactions (decisions, corrections, timing, communication style) and to read local workspace files to pre-load context. It explicitly forbids logging raw message content and states 'local only' network constraint. This is appropriate for personalization, but the policy relies on the agent/runtime to enforce 'no raw content' and 'no network exfiltration' — an instruction-only skill cannot itself enforce those protections. Also, 'pre-load workspace files' can expose sensitive local files if the workspace contains secrets; the doc does not provide technical safeguards (e.g., file path allowlist) beyond the high-level promise.
Install Mechanism
No install spec and no code files that execute on install — lowest-risk distribution model. All behavior would come from agent runtime following SKILL.md instructions.
Credentials
The skill requests no environment variables or external credentials (proportionate). However, it requires access to operator messages and local workspace files to function. Those accesses are expected given the purpose but can be sensitive; the skill's privacy principles (no raw content, operator-owned model, ability to reset/inspect) mitigate risk conceptually but are not enforceable here.
Persistence & Privilege
always is false and there are no special persistence or system-config modifications requested. The model is stored under imprint/ in the workspace — a scoped, non-system location. Autonomous invocation is allowed (platform default) but not combined with other high-risk privileges.
Assessment
This skill appears coherent with its stated purpose but depends entirely on the agent runtime honoring its privacy rules. Before installing, confirm: (1) where the agent will store imprint/operator-model.json and who can read it; (2) that the agent enforces 'no raw message logging' and 'no network calls' (consider sandboxing or network restrictions); (3) which workspace paths the agent can read (use an allowlist or dedicated directory to avoid accidental access to secrets); (4) that you can inspect, edit, reset, and delete the model easily; and (5) how multi-operator contexts are handled. If you cannot verify these runtime protections, test with non-sensitive data first and consider restricting the skill's file-read scope or running it in a confined workspace.Like a lobster shell, security has layers — review code before you run it.
agent-intelligencevk971fmry50q2qta05kfqz2t4k183a45dlatestvk971fmry50q2qta05kfqz2t4k183a45dlearningvk971fmry50q2qta05kfqz2t4k183a45dnext-genvk971fmry50q2qta05kfqz2t4k183a45doperator-modelingvk971fmry50q2qta05kfqz2t4k183a45dpersonalizationvk971fmry50q2qta05kfqz2t4k183a45d
License
MIT-0
Free to use, modify, and redistribute. No attribution required.