All Claude Code modsucsandman/Agnostic-AI
spike-describe
Spike: does a tool.describe rewrite reach the model, does agent.offer hide a type, does prompt.suggest fire?
spike-describe reaches L2 (writes or runs) and passed claude plugin validate on Claude Code 2.1.302.
- Events hooked: 4
- API methods referenced: 2
- Access from classified calls: L2, writes or runs: writes files, writes the prompt box
- Static scan, not a runtime test
To install it, open the repository and follow the author’s setup instructions. Mods need Claude Code 2.1.287 or later.
Access
L2 Writes or runswrites files, writes the prompt boxValidation
Passed on Claude Code 2.1.302View the source on GitHub Back to the list
What the scan recorded
- Observed events
- Only what it hooks
- Hooks
tool.describe {tool=Bash}, agent.offer {agent=Plan}, agent.offer, prompt.suggest, turn.complete- API calls
$.fs.write, $.prompt.suggest- Validation on 2.1.302
- Passed
- Scanned revision
aab09ff0d953
These are the hooks and API calls reported by claude plugin validate. Access levels describe the widest reach of those calls and are not safety ratings. Validation is a static check, not a runtime test, and a passing result does not prove that a mod works or is safe. How the scan works.
More from ucsandman/Agnostic-AI
- benderREALITY BENDER: intercepts a harmless tool result and alters what Claude receives; prints REAL vs RECEIVED
- blackboxBLACK BOX RECORDER: structured per-session timeline of prompts, model steps, tool calls, subagents and denial…
- claude-runtimeShared Claude runtime adapter: normalises function-hook events into ClaudeRuntimeEvent, exposes $.runtime (em…
- costclaw-liveCOSTCLAW LIVE: per-request cost, cache decay and repeated-read detection while the session runs, and it serve…
- guardianGUARDIAN: executable policy that protects lab/demos/protected/** from Write/Edit/Bash, whatever CLAUDE.md says
- harness-modsHarness Mod layer: heartbeat/canary, native usage, rewrite explanations, measured subagent accounting, superv…
- matrixMATRIX: freezes a harmless tool call before it runs, shows what Claude intended, asks you, then continues or…
- probe-uiInteractive UI probe: ui.render components, toast, status, log, pane, command register
- probe2Interception probe: deny, rewrite, replace result, delay, throw, register tool, classify
- probe3Permission override, subagent visibility, agent.spawn model rewrite, turn.step routing, prompt context inject…
- prodguardPRODUCTION GUARD: ambient project/environment context + a transport-independent policy core (DashClaw evidenc…
- spike-compactSpike: does $.session.compact() work from a plugin, and what does the session.compact event carry?
- spike-handoffSpike: safe-boundary handoff (bundle written at turn.complete from the runtime bus) and $.turn.abort semantics
- supervisorSUBAGENT SUPERVISOR: routing-and-budget Mod. Rewrites a spawn's model onto the capability graph instead of de…
- tunnel-costMIDDLEWARE TUNNEL stage 2: Cost analyzer: estimates the call's token cost from the args size
- tunnel-observerMIDDLEWARE TUNNEL stage 1: Observer: sees the call first, stamps t0, prints what enters
- tunnel-policyMIDDLEWARE TUNNEL stage 3: Policy: allows everything harmless; denies a matching rule
- tunnel-telemetryMIDDLEWARE TUNNEL stage 4: Telemetry: records the call and its result for the ledger
- xrayCLAUDE RUNTIME X-RAY: wildcard inspector with filters (/xray all|tools|subagents|session|usage|ui|errors|unkn…
Badge for your README
Authors can show the latest scan result in their README. Each badge links to this page.
[](https://mods.aidojo.si/mods/ucsandman--Agnostic-AI--spike-describe/)
[](https://mods.aidojo.si/mods/ucsandman--Agnostic-AI--spike-describe/)