registry / @johnny0120/opencode-user-predictor / 0.2.0

@johnny0120/opencode-user-predictor@0.2.0

OpenCode plugin that predicts your next message and shows it as ghost text — role-plays as you using your conversation history. /pred-on to enable.

AI Security Review

scanned 1h ago · by lpm-firewall-ai

LPM treats this as warn-only first-party agent extension lifecycle risk. At OpenCode plugin runtime, the package registers a package-owned predictor agent and commands. Explicit profile/seeding commands collect OpenCode conversation text into local project data files; prediction submits session text through the configured OpenCode client/model.

Static reason

No blocking static signals were detected.

Trigger

Loading the OpenCode plugin enables its hooks; `/pred-seed` or `/pred-profile` activates local history collection.

Impact

Conversation content can be retained in local `data/` files and supplied to the user-configured model provider; no package-controlled exfiltration is confirmed.

Mechanism

OpenCode extension registration with user-history ingestion and model-backed prediction.

Rationale

No concrete malicious chain is present, but the package-owned OpenCode extension can collect and retain user conversation history and invoke the configured model. Flag as a non-blocking lifecycle/capability warning rather than clean under the stated extension policy.

Evidence

package.jsonopencode.jsonsrc/index.tssrc/seed.tssrc/prompts.tspredictor-profile.jsondata/predictor.logdata/user-corpus.jsondata/user-profile-raw.json~/.local/share/opencode/opencode.db

@johnny0120/opencode-user-predictor@0.2.0

AI Security Review

Decision evidence

Source & flagged code

Source & flagged code

Findings

Source & flagged code