Back to skill

Security audit

Renesas Search

Security checks for vulnerabilities and agentic risk

Overview

This skill appears benign: it helps answer Renesas parts questions from official datasheets and saves local PDF copies, with no evidence of hidden persistence, credential access, or destructive behavior.

Install only if you are comfortable with the skill making web requests to Renesas-related pages and saving datasheet PDFs under your current workspace. Check that downloaded URLs are official Renesas links, and be aware the output template has stale ./workspace path examples even though the intended save location is embedded_dev/renesas/datasheet.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Rogue AgentSelf-Modification, Session Persistence
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (12)

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · HERMES_INSTALL.md (reported line 9)May include surrounding context.

md
| Platform | Skill format | Status |
|----------|--------------|--------|
| **OpenClaw** | `~/.openclaw/workspace/skills/renesas-search/SKILL.md` | ✅ Native |
| **ClawHub** | (any OpenClaw user) | ✅ Published as `renesas-search` v3.0.4 |
| **Hermes Agent** | `~/.hermes/skills/<category>/<name>/SKILL.md` | ✅ Adapter below |
| **agentskills.io** (Anthropic, NVIDIA, …) | local install of SKILL.md | ✅ Compatible |

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · TEST_SCENARIOS.md (reported line 266)May include surrounding context.

md
| Platform | Skill format | Status |
|----------|--------------|--------|
| **OpenClaw** | `~/.openclaw/workspace/skills/renesas-search/SKILL.md` | ✅ Native |
| **ClawHub** | (any OpenClaw user) | ✅ Published as `renesas-search` v3.0.4 |
| **Hermes Agent** | `~/.hermes/skills/<category>/<name>/SKILL.md` | ✅ Adapter below |
| **agentskills.io** (Anthropic, NVIDIA, …) | local install of SKILL.md | ✅ Compatible |

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · HERMES_INSTALL.md (reported line 74)May include surrounding context.

bash
git clone https://github.com/wangp-gh/renesas-search.git /tmp/renesas-search
mkdir -p ~/.hermes/skills/hardware
cp -r /tmp/renesas-search ~/.hermes/skills/hardware/renesas-search
# Copy only the skill files you want:
cp ~/.hermes/skills/hardware/renesas-search/SKILL.md ~/.hermes/skills/hardware/renesas-search/SKILL.md

Unrestricted Tool Access

Medium
Category
Excessive Agency
Confidence
80% confidence
Finding

Skill grants unrestricted tool access without appropriate constraints. An agent with unfettered tool access can perform arbitrary actions including file modification, network requests, and code execution.

Content

Scanner excerpt · HERMES_INSTALL.md (reported line 129)May include surrounding context.

md
## Notes on the Hermes Security Scan

Hermes' `Curator` system sandboxes skills installed from external sources (Hub, GitHub, etc.) for safety. `renesas-search` is a read-only documentation skill — it only fetches datasheets and renders tables — so it should pass the scan without issues. The skill does **not** execute arbitrary code, write outside `embedded_dev/`, or modify the agent's core behavior.

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · README.md (reported line 47)May include surrounding context.

md
Many Renesas datasheets publish **Typ-only values** in their electrical characteristics tables — the Min and Max columns are blank. When citing such a value:

- ✅ Write `Typ: 6.8 µA` (qualified) and a note "Min/Max: not specified"
- ❌ Do NOT write `6.8 µA` alone as if it were a full spec

## Installation

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · README.md (reported line 47)May include surrounding context.

md
Many Renesas datasheets publish **Typ-only values** in their electrical characteristics tables — the Min and Max columns are blank. When citing such a value:

- ✅ Write `Typ: 6.8 μA` (qualified) and a note "Min/Max: not specified"
- ❌ Do NOT write `6.8 μA` alone as if it were a full spec

## Installation

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The output template points to ./workspace/<PartNumber>_datasheet.pdf, which contradicts the earlier mandatory save location ./embedded_dev/renesas/datasheet/. This inconsistency can cause downstream agents or users to read from the wrong path, fail to find the downloaded file, or accidentally trust a stale/local file from an unintended directory.

Content

No source excerpt is available for this finding.

Scope Creep

Low
Category
Excessive Agency
Confidence
70% confidence
Finding

Skill's behavior or capabilities extend beyond its stated purpose. Scope creep allows an agent to perform actions unrelated to its documented functionality, increasing the attack surface.

Content

Scanner excerpt · README.md (reported line 10)May include surrounding context.

md
This skill triggers ONLY when the user explicitly mentions:

- **"Renesas"** / **"瑞萨"**
- A Renesas product family name, including (but not limited to):
  - **BLE SoC**: DA14531, DA14592, DA14594, DA14697, DA1470x
  - **MCU**: RA, RA2M1, RA4M2, RA4M3, RA6M2, RA6M3, RA6M4, RA6M5
  - **Low-power MCU**: RL78/G14, RL78/G23, RL78/H1D, RL78/I1D

Scope Creep

Low
Category
Excessive Agency
Confidence
70% confidence
Finding

Skill's behavior or capabilities extend beyond its stated purpose. Scope creep allows an agent to perform actions unrelated to its documented functionality, increasing the attack surface.

Content

Scanner excerpt · SKILL.md (reported line 64)May include surrounding context.

md
This skill triggers ONLY when the user explicitly mentions:

- **"Renesas"** / **"瑞萨"**
- A Renesas product family name, including (but not limited to):
  - **BLE SoC**: DA14531, DA14592, DA14594, DA14697, DA1470x
  - **MCU**: RA, RA2M1, RA4M2, RA4M3, RA6M2, RA6M3, RA6M4, RA6M5
  - **Low-power MCU**: RL78/G14, RL78/G23, RL78/H1D, RL78/I1D

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
92% confidence
Finding

The skill mandates downloading PDFs into the current working directory without requiring explicit user awareness or confirmation that the workspace will be modified. Silent filesystem writes and network access are risky because they create side effects beyond answering a question, especially in shared or sensitive workspaces.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
90% confidence
Finding

The workflow includes mkdir, curl, file, and other shell-driven actions that create directories and download remote content, but it does not require a safety notice or confirmation before performing those side effects. If invoked automatically, this can surprise users, modify repositories/workspaces, and introduce unreviewed external files into the environment.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
92% confidence
Finding

The test scenarios explicitly instruct the agent to download datasheet PDFs into the current working directory, causing local file writes as part of normal operation. While the files are from an official vendor source and the write path is constrained, this still creates side effects without an explicit user-consent step and could surprise operators or clutter/overwrite local workspace contents.

Content

No source excerpt is available for this finding.