Back to skill

Security audit

skill-authoring

Security checks for vulnerabilities and agentic risk

Overview

ClawScan marks this skill suspicious with medium confidence, mainly because scanners report unsafe deployment guidance and privacy risks in testing advice, while its stated skill-authoring purpose is otherwise coherent.

Consider the verdict’s Review status before installing. Pay particular attention to the reported risk of overwriting an existing skill during deployment and the advice to reuse real prompts or transcripts without a stated sanitization step. Skillspector’s analysis is partial: it reports missing referenced paths, so the evidence is incomplete.

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
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Memory PoisoningPersistent Context Injection, Context Window Stuffing, Memory Manipulation
  • Rogue AgentSelf-Modification, Session Persistence
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (15)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
85% confidence
Finding

Authenticated network requests are not part of the declared skill-authoring purpose, and the snippet provides no connection to TDD or skill validation. Its location in a Markdown file leaves open the possibility that it is merely an example, but the supplied content does not establish that context. The incomplete request does not establish a destination, successful execution, or malicious exfiltration.

Content

No source excerpt is available for this finding.

Memory Manipulation

High
Category
Memory Poisoning
Confidence
90% confidence
Finding

Skill manipulates agent memory, state, or stored context. Memory corruption can alter personality, override safety rules, or cause unpredictable behavior.

Content

Scanner excerpt · modules/error-handling.md (reported line 201)May include surrounding context.

md
For idempotent reads (a `gh api` GET), one retry is reasonable.
For state-changing operations (commits, pushes, file writes),
silent retries cause duplicate work or corrupt state. Default
to no retry. Add retries only for specific commands with
documented idempotence.

Self-Modification

High
Category
Rogue Agent
Confidence
85% confidence
Finding

Skill modifies its own code, configuration, or behavior at runtime. Self-modification enables an agent to escalate privileges, disable safety constraints, or install persistent backdoors.

Content

Scanner excerpt · modules/progressive-disclosure.md (reported line 170)May include surrounding context.

md
## Example: Secure API Skill

1. Document baseline (3 scenarios)
2. Write skill addressing failures
3. Add anti-rationalization

For additional examples, see `modules/examples.md`

Self-Modification

High
Category
Rogue Agent
Confidence
85% confidence
Finding

Skill modifies its own code, configuration, or behavior at runtime. Self-modification enables an agent to escalate privileges, disable safety constraints, or install persistent backdoors.

Content

Scanner excerpt · modules/progressive-disclosure.md (reported line 277)May include surrounding context.

md
## Quick Start

1. Document failures
2. Write skill
3. Test improvements

See modules for details.

Self-Modification

High
Category
Rogue Agent
Confidence
85% confidence
Finding

Skill modifies its own code, configuration, or behavior at runtime. Self-modification enables an agent to escalate privileges, disable safety constraints, or install persistent backdoors.

Content

Scanner excerpt · modules/tdd-methodology.md (reported line 447)May include surrounding context.

Writing Skill First

text
1. Write skill based on theory
2. Hope it works
3. Maybe test later

Deployment copies into the active skills directory without warning that existing skill files may be overwritten.

Medium
Category
Output Handling
Confidence
90% confidence
Finding

The deployment example recursively copies a skill into the active skills directory without checking whether the destination already exists. On a name collision, this can overwrite existing instructions and leave stale files behind, unintentionally changing behavior across subsequent sessions; deployment is legitimate here, but the missing collision safeguard is unsafe.

Content

No source excerpt is available for this finding.

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 · modules/examples.md (reported line 29)May include surrounding context.

md
**Skill**: `Skill(pensive:tiered-audit)`
**Path**:
`plugins/pensive/skills/tiered-audit/SKILL.md`

What to copy from this skill:

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 · modules/examples.md (reported line 56)May include surrounding context.

md
**Skill**: `Skill(abstract:subagent-testing)`
**Path**:
`plugins/abstract/skills/subagent-testing/SKILL.md`

This skill exists because `skill-authoring` requires baseline
tests in fresh subagents. Rather than restating the testing

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 · modules/examples.md (reported line 87)May include surrounding context.

md
**Skill**: `Skill(imbue:proof-of-work)`
**Path**:
`plugins/imbue/skills/proof-of-work/SKILL.md`

This skill enforces evidence before "done." It is referenced
by `skill-authoring/SKILL.md` as the source of detailed

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 · modules/examples.md (reported line 120)May include surrounding context.

md
**Skill**: `Skill(scribe:slop-detector)`
**Path**:
`plugins/scribe/skills/slop-detector/SKILL.md`

What to copy:

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 · modules/examples.md (reported line 141)May include surrounding context.

md
**Skill**: `Skill(abstract:skill-graph-audit)`
**Path**:
`plugins/abstract/skills/skill-graph-audit/SKILL.md`

What to copy:

Real session transcripts are reused and persisted as test fixtures without a privacy warning or sanitization step.

Medium
Category
Output Handling
Confidence
93% confidence
Finding

The module recommends drawing test prompts from real session transcripts while also directing authors to persist prompts and full responses in repository fixtures. Without a sanitization or authorization gate, this can copy credentials, personal information, or confidential project data into durable files and additional subagent contexts.

Content

No source excerpt is available for this finding.

Real user prompts are reused in testing without a privacy warning or sanitization step.

Medium
Category
Output Handling
Confidence
90% confidence
Finding

The troubleshooting guidance explicitly recommends replaying an exact real-user prompt through a fresh subagent and adding it to the test corpus, without checking for secrets, personal information, or authorization to reuse it. This can unnecessarily expose sensitive content to additional agent contexts or persist it in test fixtures and repository history, although the file does not establish any external exfiltration destination.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · modules/troubleshooting.md (reported line 230)May include surrounding context.

md
2. **Skill instructs Claude to run a command without
   capturing output**. Claude infers what the output would
   have been.
3. **No verification step**. The skill produces an artifact
   without checking it.

### Diagnosis

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · modules/troubleshooting.md (reported line 231)May include surrounding context.

md
capturing output**. Claude infers what the output would
   have been.
3. **No verification step**. The skill produces an artifact
   without checking it.

### Diagnosis