Security audit
Security checks for vulnerabilities and agentic risk
This is a markdown-only reasoning skill that asks the agent to pause and think through failure modes before consequential actions.
Installers should understand that this skill may cause the agent to apply an extra reasoning checklist on many non-trivial tasks, which can add verbosity or small delays, but the reviewed artifact does not add tools, persistence, credentials, network access, or hidden execution behavior.
Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).
## Scenario 3: The Confident Wrong Answer
**Task**: User asks "Is it safe to run `chmod 777` on my web server's upload
directory?"
### Without Inversion Protocol
The description says to use the skill when the agent is about to execute code, write files, run commands, answer complex questions, debug issues, make architectural decisions, or perform any action where being wrong has consequences. This trigger scope is extremely expansive and overlaps with common, everyday agent behavior, making it unclear when the skill should not activate beyond a few examples later in the file.
The phrase "before any action where being wrong matters" and examples like "performing multi-step workflows" or situations where "the user has corrected you before" are subjective and open to broad interpretation. Although a few trivial exclusions are listed, the file still lacks sufficiently specific constraints to distinguish routine from non-routine cases consistently.
Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.
## Scenario 3: The Confident Wrong Answer
**Task**: User asks "Is it safe to run `chmod 777` on my web server's upload
directory?"
### Without Inversion Protocol
Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.
## Scenario 3: The Confident Wrong Answer
**Task**: User asks "Is it safe to run `chmod 777` on my web server's upload
directory?"
### Without Inversion Protocol
Tool defaults are unsafe or overly permissive (e.g. disabled TLS verification, no authentication, world-writable permissions). Unsafe defaults widen the attack surface.
### Without Inversion Protocol
Agent explains what chmod 777 does and warns it's "generally not recommended"
but provides the command. User runs it. Attacker exploits the world-writable
directory within hours.
### With Inversion Protocol
Tool defaults are unsafe or overly permissive (e.g. disabled TLS verification, no authentication, world-writable permissions). Unsafe defaults widen the attack surface.
### Without Inversion Protocol
Agent explains what chmod 777 does and warns it's "generally not recommended"
but provides the command. User runs it. Attacker exploits the world-writable
directory within hours.
### With Inversion Protocol
Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.
→ Adjust: Explain the risk clearly, provide the SAFE permission set instead
**Result**: Agent provides `chmod 750` with www-data group ownership —
functionally equivalent for the user's need, without the security hole.
---