- Home
- Skills
- Agents & Orchestration
- Ghost Busters — Silent Failure Hunter
Works with the AI Agents you already use
Ghost Busters — Silent Failure Hunter
A systematic protocol for hunting silent production bugs, stalled pipelines, and data mismatches in ERP/CRM systems.
Free
Ghost Busters — Silent Failure Hunter
Example session with this skill installed
Our CRM synchronization reports SUCCESS, but no new orders have appeared for 48 hours. The source data hasn't changed, while our accounting system shows new activity. Use Ghost Busters to investigate safely without modifying production.
- Read your context and instructions
- Compiled the ghost busters
Ghost Busters — Silent Failure Investigation
SIMULATED DEMO — synthetic data
Investigation Summary
The synchronization reports successful executions, but there is no evidence of new data being imported.
| Evidence | Observation |
|---|---|
| Successful sync runs | 96 |
| Imported records | 0 |
| Identical source snapshots | 5 |
| New CRM orders | 0 |
| Independent activity | 84 documents |
Ghost Pulse Assessment
Status: SUSPECTED_SILENT_FAILURE
Priority: P1
A successful scheduled task does not prove the data pipeline is functioning. Independent business activity combined with unchanged source data indicates a potential silent failure.
Evidence Gates
- Environment: Verification required
- Pulse: Suspected silent failure
- Impact: Not evaluated
- Release: Not evaluated
Recommended Investigation
- Confirm the correct production environment.
- Check data freshness, source ownership, and business hours.
- Compare CRM records with the independent accounting source.
- Trace the pipeline to identify where progress stops.
- Reproduce the defect with a regression test.
- Verify any proposed repair before deployment.
Conclusion
A silent failure is suspected. The root cause is not yet proven.
No production data was modified.
Next step: collect independent read-only evidence and begin the eight-phase Ghost Busters investigation.
Connects securely to your tools. The creator never sees your data.
About this skill
Ghost Busters is a systematic debugging skill designed to eliminate "ghosts"—bugs that fail silently without throwing exceptions. It addresses critical production failures like stalled pipelines, empty screens that show success instead of errors, and data mismatches between systems. The tool moves beyond reading code to analyzing production data, logs, and deployment history through a rigorous eight-phase hunt protocol.
What it does
- Identifies silent failures by detecting frozen syncs, stalled cron jobs, and empty results masked by row-level security.
- Verifies environment identity to ensure investigations target the actual database used by the deployed application.
- Assesses fix impact by calculating how widening filters or removing gates affects bulk actions and dependencies.
- Automates class-based raids to scan the codebase for sibling bugs once a specific failure pattern is confirmed.
- Enforces evidence gates requiring verified production observations before labeling a diagnosis as proven.
How it works
- Initiate the Hunt by restating the symptom with specific numbers, screens, and date windows to define the scope.
- Consult the Archive to query engineering memory and past incident postmortems before forming any hypothesis.
- Deploy Probes using SQL instruments to check pipeline liveness, snapshot completeness, and role-based permissions.
- Execute Exorcism by documenting the rollout order, compatibility plan, and recovery steps before applying changes.
Frameworks & tools
This skill is optimized for React and Supabase stacks (PostgreSQL), utilizing supabase-js, custom AST scanners, and offline evidence evaluators via Node.js scripts.
Why this beats prompting it yourself
Standard prompts often guess root causes based on code logic alone, which misses silent production state issues. This skill enforces a non-negotiable protocol that prevents "fixed one, broke three" regressions by requiring impact analysis and watchdog verification.
Use cases
- Investigating why a CRM total does not match an accounting system despite no errors.
- Debugging a "success" status on a sync process that hasn't actually updated records in days.
- Sweeping a module for unbounded reads or row-cap truncations after catching one instance.
Known limitations
Requires operator-prepared aggregate JSON for evidence gates. It does not connect directly to production or perform automated writes to the database.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Agents & Orchestration.
Requires operator-prepared aggregate JSON for gates. Cannot directly write to production. Optimized for Supabase/PostgreSQL stacks. RLS evidence requires specific role impersonation.
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 buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
1 use
Times used to date
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 25 days ago
- Free with an Agensi account
- Software
Node.js 20+
- Optional
Git
Permissions
Ghost Busters uses terminal access to run diagnostic scripts, file reading to inspect source code and evidence, and local file writing for reports, tests, and alert state. Network access and environment variables are optional, used only for explicitly configured read-only HTTPS evidence collectors. No production write access, automatic deployment, or remediation is required.