- Home
- Skills
- Agents & Orchestration
- jev advisor
Works with the AI Agents you already use
jev advisor
Internal decision-support layer for agent tool selection, risk assessment, and task completion verification.
$5
jev advisor
Example session with this skill installed
Analyze the current workspace state and recent test failures to decide if we should retry the build or refactor the database connector. Prepare the JSON request for Jev Advisor.
- Read your context and instructions
- Compiled the jev advisor
I have analyzed the recent npm test failure and the current git status. The failure appears to be a timeout in the database connection pool. I am consulting the Jev Advisor to weigh the risk of a simple retry versus a configuration refactor.
Jev Advisor Request & Recommendation
{
"context": {
"goal": "Fix flaky database connection tests",
"observed_facts": [
"Test 'Connection Pool Exhaustion' failed with ETIMEDOUT",
"Current pool size is set to 5",
"Git diff shows recent changes in pool-manager.ts"
],
"tool_registry": ["shell_execute", "file_read", "file_edit"],
"candidates": [
{
"action": "shell_execute",
"params": "npm test",
"intent": "Retry build to check for flakiness"
},
{
"action": "file_edit",
"params": "pool-manager.ts",
"intent": "Increase max pool size to 10"
}
]
},
"advisor_response": {
"recommended_action": "file_edit",
"scores": {
"information_gain": 0.4,
"success_probability": 0.85,
"token_cost": 0.2
},
"risks": ["minor_config_drift"],
"ask_user": false
}
}
The advisor recommends refactoring the pool size rather than retrying, as the ETIMEDOUT error combined with a small pool size (5) provides high evidence of a resource constraint.
Next steps
- Apply the recommended increase to the connection pool size in
pool-manager.ts. - Execute the test suite again to verify the fix.
- If tests pass, perform a final completion check with the advisor to ensure no side effects.
Connects securely to your tools. The creator never sees your data.
About this skill
Manual tool selection often leads to inefficient token usage or circular reasoning loops. Jev Advisor provides a structured decision-support layer for agents to evaluate tool candidates and verify task completion before committing to an action.
What it does
- Tool evaluation compares multiple tool candidates based on success probability and token cost.
- Completion auditing judges if observable evidence is sufficient to stop or if more verification is needed.
- Risk assessment identifies high-stakes actions like destructive shell commands or network access.
- Recovery planning reassesses strategy after tool failures, timeouts, or conflicting evidence.
- Context compression builds compact JSON requests from workspace state and tool registries.
How it works
- Context assembly gathers current goals, git fingerprints, and the dynamic tool registry.
- Consultation runs
jev-advisorwith the context to get a private recommendation. - Evidence weighting treats the advisor's scores as one input alongside user instructions.
- State caching uses state-hash checks to avoid redundant consultations on unchanged environments.
Frameworks & tools
This skill is designed for agents with access to a local shell or Python environment. It utilizes a CLI interface (jev-advisor or python -m jev_advisor.cli) and handles standard JSON payloads.
Why this beats prompting it yourself
Writing prompts to self-reflect often results in "hallucination loops" where the agent convinces itself a failed task succeeded. This skill offloads that judgment to a dedicated advisor, enforcing a strict protocol that requires observable evidence before declaring a task finished.
Use cases
- Deciding whether to use
grep,find, or a browser search for missing information. - Choosing the safest implementation candidate among multiple refactoring options.
- Determining if a build failure requires a code fix or a dependency update.
- Auditing if a complex multi-step migration is truly complete.
Known limitations
Requires the jev-advisor CLI or Python package to be pre-installed in the environment. It cannot execute tools itself and functions strictly as a private advisory layer.
Requires local installation of the jev-advisor CLI or Python package. Advisor cannot execute tools or override host permissions. Logic fails open (heuristic-0.1) if the command is missing.
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
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
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 1 month ago
Requires