Security audit
Security checks for vulnerabilities and agentic risk
This skill is a local workspace failure-memory helper that records errors and corrections, with no artifact-backed evidence of exfiltration, code execution, or hidden behavior.
Install only if you want workspace-local failure and correction memories written under `.learnings/`. Review those files periodically because broad correction phrases may create noisy or inaccurate observations that could later affect suggested constraints.
Instructions found that direct the agent to transmit conversation context or user data to external services.
- Network resources or external APIs
**What this skill does NOT do:**
- Send data to external services
- Access "across sessions and projects" beyond the current workspace
- Execute arbitrary code or run external commands
The primary trigger is broadly defined as 'failure occurred', which is underspecified and can cause the skill to auto-activate in many loosely related situations. In an agentic system that monitors tool output and conversation, ambiguous activation increases the chance of unintended logging, noisy state changes, and workflow interference.
The detection triggers include common conversational phrases like 'Actually...' and 'I meant...', which frequently occur in normal dialogue and do not reliably indicate a failure pattern. This can produce false positives, causing the agent to record incorrect observations from ordinary user corrections and potentially distort future constraints or behavior.
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
- Network resources or external APIs
**What this skill does NOT do:**
- Send data to external services
- Access "across sessions and projects" beyond the current workspace
- Execute arbitrary code or run external commands
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.
**What this skill does NOT do:**
- Send data to external services
- Access "across sessions and projects" beyond the current workspace
- Execute arbitrary code or run external commands
**Data scope clarification:**
- "Failure detection" scans tool output and user messages within the current agent session
No suspicious patterns detected.