How do I install the Reveng MCP Annotation Honesty skill?

Run npx skillmds@latest add oimiragieo/reveng-mcp-annotation-honesty in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Reveng MCP Annotation Honesty safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Reveng MCP Annotation Honesty?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Reveng MCP Annotation Honesty free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Reveng MCP Annotation Honesty?

oimiragieo (@oimiragieo) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Reveng MCP Annotation Honesty by oimiragieo

Name: Reveng MCP Annotation Honesty
Author: oimiragieo

Reveng MCP Annotation Honesty

States REVENG Wave 2 MCP tool-annotation honesty for denylist dual-labels and actlint expectations (L45). Use when editing MCP tool annotations, risk_level hints, generate_exploit / recompile_binary labels, denylist policy, or when tempted to treat annotations as enforcement or auto-map every high risk tool.

oimiragieo Updated Sep 17, 2026 0 repo stars

File contents

REVENG MCP annotation honesty (Wave 2 / L45)

Overview

MCP tool annotations are advisory hints, not a sandbox or security boundary. Wave 2 honesty is an explicit denylist of dual-labels — not risk_level == "high" auto-mapping.

Rules

Hints ≠ enforcement — clients/hosts must not treat annotations as authorization. Cite MCP blog (see references).
Wave 2 denylist only (dual-label both MCP hints + proprietary keys):
- generate_exploit
- recompile_binary
Do not auto-map every risk_level == "high" tool to destructiveHint (e.g. analyze_memory_dump, ai_code_reconstruction wait for later waves).
CI direction — eventual tools/list parity with annotation tests (sunpeak / Closient patterns). Full actlint CI = future wave — do not claim Wave 2 closed actlint product GA.
Policy + tests (canonical):
- docs/architecture/policy-mcp-annotation-honesty-wave2.md
- tests/unit/test_mcp_annotation_honesty_wave2.py

Anti-patterns

Excuse	Reality
"Annotations are set — tools are sandboxed"	Annotations are hints only (MCP blog).
"All high-risk tools should get destructiveHint now"	Wave 2 = explicit denylist (L45), not auto-map.
"Ship full actlint in this PR — honesty incomplete otherwise"	Full actlint / mcp-conform CI is deferred; don’t overclaim.
"Under-declare hints to avoid noise"	Under-declared hints are worse than over-declared (actlint severity).

Cross-refs

reveng-release-honesty (L45 one-liner)
Pins: references/pins.md

---name: reveng-mcp-annotation-honestydescription: States REVENG Wave 2 MCP tool-annotation honesty for denylist dual-labels and actlint expectations (L45). Use when editing MCP tool annotations, risk_level hints, generate_exploit / recompile_binary labels, denylist policy, or when tempted to treat annotations as enforcement or auto-map every high risk tool.---# REVENG MCP annotation honesty (Wave 2 / L45)## OverviewMCP tool annotations are **advisory hints**, not a sandbox or security boundary.Wave 2 honesty is an **explicit denylist** of dual-labels — not`risk_level == "high"` auto-mapping.## Rules1. **Hints ≠ enforcement** — clients/hosts must not treat annotations as authorization. Cite MCP blog (see references).2. **Wave 2 denylist only** (dual-label both MCP hints + proprietary keys): - `generate_exploit` - `recompile_binary`3. **Do not** auto-map every `risk_level == "high"` tool to `destructiveHint` (e.g. `analyze_memory_dump`, `ai_code_reconstruction` wait for later waves).4. **CI direction** — eventual `tools/list` parity with annotation tests (sunpeak / Closient patterns). **Full actlint CI = future wave** — do not claim Wave 2 closed actlint product GA.5. Policy + tests (canonical): - `docs/architecture/policy-mcp-annotation-honesty-wave2.md` - `tests/unit/test_mcp_annotation_honesty_wave2.py`## Anti-patterns| Excuse | Reality ||--------|---------|| "Annotations are set — tools are sandboxed" | Annotations are hints only (MCP blog). || "All high-risk tools should get destructiveHint now" | Wave 2 = explicit denylist (L45), not auto-map. || "Ship full actlint in this PR — honesty incomplete otherwise" | Full actlint / mcp-conform CI is deferred; don’t overclaim. || "Under-declare hints to avoid noise" | Under-declared hints are worse than over-declared (actlint severity). |## Cross-refs- `reveng-release-honesty` (L45 one-liner)- Pins: [references/pins.md](references/pins.md)

Frequently asked questions