← Back to .knowledge

// Blog / FAQ

FAQ: .knowledge for teams

A practical FAQ for teams that want coding agents to share repo-local context without pretending summaries are source of truth.

// direct answer

Short answer

.knowledge is a repo-local trust and routing layer for AI coding agents. It helps teams share context across tools, mark stale or suspect knowledge, inspect evidence, and keep source code and tests above generated summaries.

Answer common team questions about adopting .knowledge and Pro2Pilot Inspector.

// team adoption

What teams are really asking

Teams do not only ask whether an agent can read a repo. They ask whether context survives handoff, whether stale summaries are visible, and whether reviewers can see what the agent trusted.

.knowledge answers those questions with files inside the repository instead of private chat memory.

Pro2Pilot Inspector adds the visual workflow layer when the team needs repair boards, PR impact, policies, and cross-review operations.

// operating rule

The rule to remember

Use .knowledge to route, inspect, and govern agent context.

Use current source and tests to decide behavior.

When trust is weak or freshness is stale, make the agent re-read code before making claims or changes.

// concrete example

A realistic adoption path

Start by adding .knowledge to one active repository and installing Codex, Claude Code, or OpenCode integrations.

Run import and release checks, then use the static Inspector to review trust and repair state.

Once the team sees repeated handoff or PR-review pain, add Inspector workflows on top of the open core.

// repo proof

Repo proof to inspect

Team entrypoint

.knowledge/Quick-Start.md

The first-read instructions live in the repo, not in one person's prompt notes.

Use it for
  • Team onboarding
  • Agent setup
Do not use it for
  • Behavior proof
Shared route

.knowledge/maintenance/routing_bundle.json

Every agent can start from the same compact route.

Use it for
  • Session start
  • Module discovery
Do not use it for
  • Skipping source review
Shared trust state

.knowledge/maintenance/trust_report.json

The team can see what is trusted, stale, suspect, or low-confidence.

Use it for
  • Review readiness
  • Agent guardrails
Do not use it for
  • Approving risky edits
Visual review

.knowledge/inspector/index.html

A generated Inspector makes trust and repair state easier to discuss.

Use it for
  • Team reviews
  • Repair planning
Do not use it for
  • Replacing current code

// command transcript

Commands and expected checks

node .knowledge/tools/install-agent-integrations.js
expected
Team agent entrypoints are installed or refreshed.
inspect next
.agents/skills/ .claude/skills/ and .opencode/commands/
node .knowledge/tools/flow.js import
expected
Initial repository context is generated without overwriting curated knowledge.
inspect next
.knowledge/project_index.json and .knowledge/modules/
node .knowledge/tools/flow.js release --no-color
expected
Routing, trust, freshness, metrics, search, and Inspector artifacts are rebuilt.
inspect next
.knowledge/maintenance/quality_report.json
FAQ grid repo-local proof

What teams need to decide

01 Does the repo own the context?
02 Can reviewers see trust and freshness?
03 Can multiple agents share the same first-read path?
04 Can source and tests override summaries?
05 Can unresolved context debt become visible work?

// field guide

Team adoption questions

The practical answers should be inspectable in the repository.
QuestionShort answerArtifact to inspect
Is it open core?Yes, the repo-local layer is open and local-first..knowledge/tools/ and agent-integrations/
Does it replace tests?No. It routes agents toward the right tests.source files and test files
Does it replace RAG?No. It complements retrieval with trust and freshness.trust_report.json and search/index.json
Can teams review it?Yes, because the context is stored as repo files..knowledge/ and Inspector output
When is Inspector Pro useful?When the team wants visual repair, policy, and PR impact workflows./inspector/ and waitlist

// guardrails

What the agent should not trust blindly

  • Do not sell .knowledge as agent memory that makes code review unnecessary.
  • Do not hide stale or suspect modules to make adoption look cleaner.
  • Make the source-of-truth order explicit during team onboarding.

// common mistakes

Common mistakes

  • Starting with a hosted memory discussion before fixing repo-local onboarding.
  • Treating FAQ answers as policy instead of linking them to artifacts.
  • Adding Inspector visuals before the underlying trust files are healthy enough to show.

// quick FAQ

FAQ

Is .knowledge a product or a standard?

The core is a repo-local open layer. Pro2Pilot builds product workflows, Inspector views, and team operations on top of it.

What should a team try first?

Install it in one active repository, run import and release, open the Inspector, and make one AI-generated PR include trust and source checks.

// page-specific next step

Use this page when onboarding a team

Start with one repository, install integrations, run import and release checks, and show the team how trust state changes the agent workflow.