pod
Other formats agents might prefer:
markdownjsonllms.txt

Agent? You probably want markdown or json, or Pod over MCP.

Reported issues for Doc Bridge

Pod holds 20 of 20 problems reported by people outside the maintainer team. Issues filed by the project's own owners, members and collaborators are excluded entirely — a maintainer's release checklist is not a warning to a prospective user.

Back to Doc Bridge.

Most discussed

Phase 1: MCP tools — doc.search, doc.get, handoff.resolve

Part of #1

Goal

Expose the same deterministic index via a local MCP stdio server for Cursor, Claude Desktop, Continue, etc.

Deliverables

  • ak-docs mcp subcommand
  • Tools: doc.search, doc.get, handoff.resolve
  • Zod-validated tool inputs
  • Static tool descriptor export for server registration
  • Local filesystem hardening for indexed doc reads

Acceptance criteria

  • Cursor can add server and call handoff.resolve for a package id
  • Tool outputs match CLI --agent JSON f

Read the thread · 2026-07-06 · closed · 9 comments

Phase 1: CLI — ak-docs query and search with --agent handoff

Part of #1

Goal

Ship a standalone CLI for deterministic doc lookup (no LLM required).

Deliverables

  • ak-docs query package <id> [--agent]
  • ak-docs query ownership <id> --agent
  • ak-docs search <term> [--agent]
  • ak-docs list packages|intents|changes|knowledge
  • JSON default; --text human mode
  • package bin: ak-docs

Acceptance criteria

  • --agent emits valid AgentHandoff v1 JSON
  • Exit code 1 + usage on missing index with rebuild hint
  • Documented in README

Read the thread · 2026-07-06 · closed · 9 comments

Phase 1: Doc site adapters — Fumadocs, Docusaurus, Human guide bridge

Part of #1

Goal

Connect human-facing doc sites to agent handoffs via resolvable URLs.

Deliverables

  • fumadocs adapter: read content/docs/**, meta.json, emit humanDoc paths
  • docusaurus adapter: read docs/**, sidebars.js, map slugs
  • ## Human guide link sync + validation gate plugin (human-guide-links)
  • Config example for Fumadocs-only and Docusaurus-only repos

Acceptance criteria

  • AgentHandoff.humanDoc resolves for Fumadocs-style paths using

Read the thread · 2026-07-06 · closed · 6 comments

Resumable and idempotent workflow reliability

Parent PRD

#131

What to build

Harden the internal workflow state machine and artifact store so large scans can be safely resumed, repeated, cancelled, isolated, migrated, and audited without corrupting evidence or claiming completion from partial work.

Execution mode: AFK

Acceptance criteria

  • Every stage persists its input hash, output hash, pipeline version, analyzer versions, configuration hash, status, timestamps, and artifact references.
  • A rerun with identical s

Read the thread · 2026-08-28 · closed · 3 comments

Phase 3: Memory classification and promotion pipeline

Part of #1

Goal

Integrate local agent memory exports into doc-bridge candidates, then later classify and promote them safely.

Deliverables

  • Memory adapters: cursor-rules (.cursor/rules/*.mdc, .cursor/rules/*.md) and playbook-memory (.agent-memory/**/*.md)
  • ak-docs memory ingest → MemoryCandidate v1[]
  • bootstrap-delta, session-export adapters (post-alpha)
  • Classifier Runtime agent: convention → for-agents/AGENTS.md; user-facing → Human guide; genera

Read the thread · 2026-07-06 · closed · 3 comments

Phase 2: Registry topology surface — docs-chat, knowledge-promoter, doc-curator

Part of #1

Goal

Document and later dogfood Registry agents that orchestrate doc maintenance.

Deliverables

  • Supervisor flow spec: doc-curator (classify → draft → verify)
  • Delegate topology: docs-chat, knowledge-promoter, code-review
  • Tools: doc-bridge MCP tools as Runtime tools
  • Publish doc-curator entry to Registry or compose from existing agents (post-alpha)
  • Runtime smoke against a private consumer doc diff (post-alpha/private)

Acceptance crit

Read the thread · 2026-07-06 · closed · 3 comments

Phase 1: Define AgentHandoff v1 and DocBridgeIndex schemas

Part of #1

Goal

Publish versioned, portable JSON schemas for the doc-bridge core artifacts.

Deliverables

  • AgentHandoff v1 — matches AKOS docs:internal:query --agent output (startHere, readBeforeEditing, editRoots, checks, optional humanDoc, playbookPatterns)
  • DocBridgeIndex v1 — contentHash, ecosystem properties, OKF-shaped knowledge[], capability refs, handoffs map
  • MemoryCandidate v1 — normalized memory ingest shape (source, fact, why, howToApply, `

Read the thread · 2026-07-06 · closed · 3 comments

AgentsKit Registry assistant guardrails

Parent PRD

#131

What to build

Harden Registry-agent assistance so the default ecosystem-doc-bridge-corpus-scanner and configured alternative AgentsKit Registry agents can classify, explain, and propose changes without becoming an unverified authority.

Execution mode: AFK

Acceptance criteria

  • The default agent identity is ecosystem-doc-bridge-corpus-scanner and is recorded with Registry origin and resolved version.
  • A project can configure an alternative Registry

Read the thread · 2026-08-28 · closed · 2 comments

Most recent

Study: Verification, recovery, privacy, and budget gate

Parent PRD

#147

What to build

Bind the study to the existing enterprise verification harness so that protocol, task, scenario, metric, privacy, budget, recovery, and cleanup requirements are fail-closed and auditable. The study must preserve valid evidence through interruption, prevent stale evidence reuse, and produce publication-safe aggregates without exposing the private consumer repository.

Acceptance criteria

  • The study has a project-local verification contract with ever

Read the thread · 2026-08-30 · open · 0 comments

Study: Controlled runner for two pinned models and three scenarios

Parent PRD

#147

What to build

Create the controlled scenario runner that executes each approved task with two pinned model configurations under three separate evidence scenarios: repository only, deterministic Doc Bridge, and Doc Bridge plus a bounded AgentsKit Registry agent. The runner must isolate observations, capture provenance and cost metrics, and validate real outcomes without allowing an agent to approve its own result.

Acceptance criteria

  • The runner supports the thr

Read the thread · 2026-08-30 · open · 0 comments

Enterprise end-to-end product validation

Parent PRD

#131

What to build

Create the final enterprise validation slice that installs the exact Doc Bridge artifact in agentskit-os and validates the complete applicable product contract across CLI, MCP, report UI, documentation, agent assistance, reliability, security boundaries, and measured outcomes.

Execution mode: HITL

Acceptance criteria

  • The exact packed Doc Bridge artifact is installed and its version, package hash, and source revision are recorded.
  • Real

Read the thread · 2026-08-28 · open · 0 comments

Versioned analyzer and plugin contract

Parent PRD

#131

What to build

Define the stable language-neutral analyzer and plugin contract that lets future language and framework support extend Doc Bridge without changing the common knowledge model or weakening validation guarantees.

Execution mode: AFK

Acceptance criteria

  • Define a versioned plugin contract for discovery, relation extraction, documentation extraction, diagnostics, coverage, capabilities, and analyzer metadata.
  • Each plugin declares supported l

Read the thread · 2026-08-28 · closed · 2 comments

Enterprise verification contract and applicability profiles

Parent PRD

#131

What to build

Implement the versioned verification contract and project profiles that make enterprise completion fail closed. The engine must support low-friction defaults while making every relaxation, non-applicable surface, required metric, and completion state explicit in machine-readable and human-readable output.

Execution mode: AFK

Acceptance criteria

  • Support the profiles default, strict, poc, custom, and enterprise through the versioned

Read the thread · 2026-08-28 · open · 2 comments

PRD: Enterprise hardening for Doc Bridge

PRD: Enterprise Hardening for Doc Bridge

Problem Statement

Doc Bridge already provides deterministic repository discovery, documentation discovery, relationship reconciliation, persistent workflow artifacts, CLI and MCP surfaces, Registry-agent assistance, and a read-only HTML report. The current implementation has proven that it can scan a large real repository, preserve evidence, expose documentation/code drift, reduce agent context, and render an interactive architecture view.

It is n

Read the thread · 2026-08-28 · open · 0 comments

[Slice 13] End-to-end dogfooding and release hardening

Parent PRD

#99

What to build

Create the final synthetic and real-repository proof that the complete Doc Bridge Knowledge Engine works end to end. Exercise discovery, documentation declarations, reconciliation, rules, persistence/recovery, CLI/JSON/MCP parity, offline HTML, deterministic fixes, approval and the optional Registry-agent proposal workflow.

Type

AFK

Scope and non-goals

This slice is release validation and compatibility hardening. It must not introduce a new product

Read the thread · 2026-08-27 · closed · 1 comment

[Slice 12] AgentsKit Registry proposal workflow

Parent PRD

#99

What to build

Add an explicit assisted workflow using a source-owned agent installed from the AgentsKit Registry. Create an AgentAdapter boundary, pass immutable snapshots/reports and scoped evidence, and accept only typed AgentProposalV1 results for discovery assistance, unresolved-relation investigation, stale-documentation analysis and documentation improvements.

Type

HITL

Scope and non-goals

The first implementation supports installed/source-owned AgentsKit

Read the thread · 2026-08-27 · closed · 2 comments

[Slice 11] Safe repository boundaries and partial analysis

Parent PRD

#99

What to build

Harden the engine for untrusted and large repositories. Enforce the repository root boundary, protect against symlink/path traversal, exclude secrets and generated directories by default, redact sensitive values, support configurable resource limits and cache reuse, and represent incomplete analysis explicitly without silent truncation.

Type

AFK

Scope and non-goals

This slice applies across discovery, workflow persistence and reporting. It does not

Read the thread · 2026-08-27 · closed · 2 comments

[Slice 08] MCP parity for snapshots, reports and workflows

Parent PRD

#99

What to build

Expose the canonical artifacts and workflow actions through MCP without creating a second analysis implementation. Add read-oriented tools for snapshot, report, diagnostics, relations and run state, plus proposal listing/approval actions where the underlying workflow permits them. Preserve all existing MCP tools and response contracts.

Type

AFK

Scope and non-goals

This slice owns MCP schemas, routing, authorization boundaries and parity tests. It d

Read the thread · 2026-08-27 · closed · 2 comments

Registry follow-up: add doc-bridge as the discovery/onboarding companion

Part of #1

Goal

Make doc-bridge discoverable from the AgentsKit Registry as a companion tool for repos that install agents.

Scope

  • Define a Registry entry or docs page for doc-bridge as a tool/package, not a copied agent.
  • Add install command and expected artifacts: doc-bridge.config.*, .doc-bridge/index.json, llms.txt, MCP tools.
  • Connect it to docs-chat, knowledge-promoter, and doc-curator topology without requiring those agents.
  • Coordinate with #8 so Registry `llms.

Read the thread · 2026-07-09 · open · 0 comments

Docs polish: getting started, examples, MCP, and positioning pages

Part of #1

Goal

Create a small docs pack that makes doc-bridge easy to adopt without reading the full spec.

Scope

  • docs/getting-started.md: install, init, index, query, ask, MCP.
  • docs/why-doc-bridge.md: problem, positioning, when not to use it.
  • docs/examples.md: pnpm monorepo, Fumadocs, Docusaurus, plain markdown.
  • docs/mcp.md: Cursor/Claude/Codex style setup notes.
  • Link these from README.

Acceptance criteria

  • A developer can choose the right config example in one pa

Read the thread · 2026-07-09 · open · 0 comments

The remaining reports are on the project's issue tracker.