on-project-start

name	on-project-start
description	Use on first entry to a new repository to run environment scanning and ask targeted boundary questions before implementation.

On Project Start

Use this skill only when entering a repository for the first time in a session, or when project boundaries are still unclear — it converts unknown project boundaries into explicit, confirmed constraints before implementation starts. For returning sessions where boundaries are already confirmed, use the Session Resume Protocol in skills/memory-and-state/SKILL.md instead.

Protocol

1. Scan repository signals

   • Inspect top-level files and dependency manifests (e.g., package.json, pom.xml, go.mod, requirements.txt, Cargo.toml, Gemfile).
   • Detect likely stack and runtime (for example Spring Boot, Django, React, Terraform, CDK).
   • Detect validation commands from CI config (.github/workflows/, Makefile, Justfile, scripts).
   • Check for existing conventions docs (CONVENTIONS.md, ARCHITECTURE.md, .editorconfig).

2. Ask targeted boundary questions

   First, ask about the budget profile — this affects every subsequent decision about which skills and rules are loaded:

   "Which token budget profile fits your setup?

   • nano — < 3,000 total tokens. Single-file Small fixes only. No skills loaded. Best for: extreme token limits, simple patches.
   • minimal — < 16K context. Small tasks only. Loads demand-triage + repo-exploration. Best for: solo devs, tight budgets.
   • standard — 16K–32K context. Small/Medium tasks. Full skill set. Best for: typical team usage. (default)
   • full — 32K+ context. All tasks including Large. All skills. Best for: large teams, complex projects."

   Skip this question only if prompt-budget.yml already has budget.profile explicitly set to a non-default value.

   Then ask only high-value technical questions that can change implementation decisions:

   • Ask only questions whose answers are not already documented in existing project files.
   • Prefer either/or questions when possible.
   • Examples:
     • "I detected Spring Boot. Is there a package naming standard I must follow?"
     • "I detected AWS IaC files. Should infrastructure changes use Terraform or CDK?"
     • "I see no test framework configured. Which testing approach should I use?"

3. Confirm and persist

   • Write the selected budget profile to prompt-budget.yml → budget.profile (create the file from docs/prompt-budget-examples.md if it does not exist yet).
   • Summarize confirmed boundaries in concise bullets.
   • Add or update project/project-manifest.md with the confirmed constraints.
   • If a new durable architectural constraint is discovered, record it according to prompt-budget.yml -> decision_log.policy.

4. Continue with normal workflow

   • Run repo-exploration then demand-triage.
   • Do not start implementation until initialization questions are resolved or explicitly deferred by user instruction.

Output format

Initialization summary:
- Budget profile selected: [nano | minimal | standard | full] — written to prompt-budget.yml
- Detected stack:
- Boundary questions asked:
- User confirmations:
- Constraints recorded in project manifest:
- Open items (if any):

Conformance self-check

• Budget profile was confirmed with user and written to prompt-budget.yml
• Stack signals were discovered from real files, not guessed
• At least one boundary question was asked when constraints were unclear
• Confirmed constraints were written to project context
• Workflow proceeded to discovery/triage before implementation

Use this skill when

• Entering a repository for the first time in a session
• Project boundaries are unclear and need explicit confirmation
• prompt-budget.yml does not yet have budget.profile set