adversarial-review-loop

    Implement an agreed plan phase by phase, with an adversarial review of each phase before moving on.

    Rank#-3200
    leinardiclaudemodsLast scanned Nov 16, 2026, 08:01 PMhttps://github.com/leinardi/adversarial-review-loop
    Created
    3 months ago
    Last commit
    yesterday
    Latest release
    adversarial-review-loop--v3.0.12 months ago

    Ecosystem

    1 related extension

    Security Findings

    5

    In-process hooks module

    SuspiciousScanner

    Runs in the Claude Code process. Hooks: agent.offer, classic.DirectoryAdded, command.run, session.start, tool.call, tool.check, turn.complete, turn.start, turn...., plus 2 event names that could not be read. Calls: $.env.get, $.fs.read, $.fs.write, $.process.run, $.process.spawn. Informational, does not affect the trust score

    hooks/mod.ts

    Hooks module intercepts agent events

    MaliciousScanner

    Hooks an event name that could not be read: treated as reaching security hooks or telemetry

    hooks/mod.ts

    Hooks module intercepts agent events3×

    SuspiciousScanner

    Hooks classic.DirectoryAdded: can answer or skip settings hooks for these events

    Hooks tool.check: can override permission decisions, a PreToolUse deny included

    Hooks agent.offer, command.run, tool.call, turn.step: can rewrite or answer model input, tool calls and messages

    hooks/mod.ts

    Contents

    1 skill bundled by this extension

    1 no risk known

    adversarial-review

    skill

    None known
    .agents/skills/adversarial-reviewd00fc08193ec…

    Editions

    4
    • Scanned Nov 16, 2026, 08:01 PMHigh
    • Scanned Nov 14, 2026, 08:01 PMHigh
    • Scanned Nov 13, 2026, 08:32 PMLow
    • Scanned Nov 10, 2026, 08:19 PMLow