More screenshots
Works with the AI Agents you already use
Claude Code
Cursor
Codex CLI
Muse
Dots+21 moreClaude CodeCursorCodex CLIMuseDots
Manus
Grok
Grok Bot+18 agentsEmberDiff
Deterministic, multi-tier risk analysis for code diffs and PRs with zero-invention enforcement.
$5.99
EmberDiff
Example session with this skill installed
Analyze this diff for risk using strict sensitivity:
--- a/auth.py
+++ b/auth.py
- if user.is_authenticated and token.is_valid():
- if user.is_authenticated:
return access_granted()
- Read your context and instructions
- Compiled the emberdiff
Analysis complete. Identified a CRITICAL security regression in auth.py due to validation removal.
{
"analysis_id": "1715634281",
"config": {
"sensitivity": "strict"
},
"input_summary": {
"files_changed": ["auth.py"],
"change_types": ["modify"],
"primary_language": "python"
},
"tier1_surface": {
"inventory": "Modified auth.py to update conditional access logic."
},
"tier2_signals": [
{
"signal": "validation removal",
"location": "auth.py:3",
"severity": "CRITICAL",
"confidence": 1.0
}
],
"tier3_assessment": [
{
"finding": "Removal of token validation check simplifies authorization bypass.",
"severity": "CRITICAL",
"confidence": 1.0,
"evidence": "- if user.is_authenticated and token.is_valid():",
"blast_notes": "Directly impacts all authentication flows relying on this helper."
}
],
"tier4_uncertainty": [],
"context_requests": [],
"tier5_remediation": [
{
"finding_ref": "Removal of token validation",
"confidence": 1.0,
"patch": "if user.is_authenticated and token.is_valid():",
"context_status": "FULLY_CONTAINED"
}
],
"tier6_cumulative": {
"highest_severity": "CRITICAL",
"contradictions": ["LIMITATION: Single unified diff provided. Cross-commit contradiction detection unavailable."]
},
"telemetry": {
"zero_invention_check": "PASS",
"missing_data_policy": "EXPLICIT_NULL",
"drift_status": "NEGATED",
"sensitivity_filter_applied": "strict"
}
}
Next steps
- Revert the removal of
token.is_valid()inauth.py. - Verify why this validation was targeted for removal in the PR description.
- Run authentication integration tests to ensure token lifecycle is correctly handled.
Connects securely to your tools. The creator never sees your data.
About this skill
The problem
Manual code reviews are prone to human fatigue, subjective bias, and missed security signals. Developers often struggle to identify the true blast radius of a change or detect contradictions across a large commit range.
What it does
- Performs multi-tier risk gating to categorize changes by impact and severity.
- Enforces a deterministic severity matrix for CRITICAL, MODERATE, and NOMINAL risks.
- Calculates confidence scores based on explicit evidence vs. implied signals.
- Identifies cross-commit contradictions where security constraints are added then later weakened.
- Generates localized, zero-invention remediation patches only when context is fully contained.
Why this beats prompting it yourself
Standard LLM prompts often "hallucinate" bugs or invent context that does not exist. This skill uses a strict zero-invention policy and a locked JSON contract, ensuring you get mechanical, evidence-bound analysis instead of conversational guesses.
Use cases
- Automated PR risk assessment to flag high-severity security regressions.
- Blast radius analysis for changes touching sensitive auth or crypto modules.
- Auditing large commit histories for contradictory logic changes.
- Filtering noise from minor refactors to focus on business-critical state mutations.
Known limitations
Two run paths for 500 lines at a time for hardcoded accuracy or a full run mode if confident.. Requires explicit file paths or function signatures for external context resolution.
Input is limited to the first 500 lines of a diff. Cross-commit contradiction detection is only available when multiple distinct diffs are provided rather than a single unified diff.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
This skill has been used 4 times.
Trust & safety
Security scan
Security scan passed 1 month ago
Permissions
No special permissions declared or detected