Argus Project Gate

IPurpose

Run the front-door Argus gate before project work begins.

This skill inspects the current repo or project, recommends the right workflow, identifies available Argus skills, and asks the operator to confirm the path before implementation.

IIGate Flow

1. Project Intake Gate

   Confirm the repo path, read local instructions, list the top-level structure, and identify validation commands.

   For a parent folder with many projects, route first to $argus-project-inventory.

   Default source-control policy: Luis projects should be Git repos unless the operator explicitly marks the work as unnecessary to version. Disposable MVP experiments may be exceptions, but the gate should still recommend Git by default.

2. Work Classification Gate

   Classify the request as skill, harness, kernel, pipeline, template, example, script, system file, package metadata, app feature, bug fix, review, release, or research.

3. Workflow Recommendation Gate

   Recommend one primary Argus workflow and any supporting skills. Prefer an existing skill before proposing a new one.

4. Operator Confirmation Gate

   Prompt the operator with the recommended workflow, selected skills, validation commands, risks, and confidence. Do not implement until the operator confirms when the request is ambiguous or high impact.

5. Execution Gate

   Proceed through the selected skill or workflow with validation after each major step.

6. Skill Gap Gate

   If no existing skill fits, complete the current project with a scoped plan, then propose a new skill or reference for future reuse.

7. Retrospective Gate

   At the end, capture only reusable lessons and recommend whether to update Argus.

IIIPrompt Pattern

Use this concise prompt when the gate needs operator input:

Argus gate check:

Recommended workflow:
[workflow]

Recommended skills:
[skills]

Validation commands:
[commands]

Risks or missing inputs:
[risks]

Confidence:
[percentage]

Proceed with this Argus path?

IVNo-Bloat Rules

• Do not load every Argus reference.
• Do not create a new skill during the main task unless explicitly requested.
• Record missing-skill opportunities for the retrospective gate.
• Keep recommendations short and actionable.
• Stop when confidence is below 95%.
• Do not add a new gate when a checklist item inside an existing gate is enough.

VOutput

Return:

• repo/project summary
• work classification
• recommended workflow
• recommended skills
• validation gates
• missing skill opportunities
• confidence
• confirmation prompt when needed