Atlas / MCP servers / benwu95 / prospec

prospecCAUTION

Name: prospec
Author: benwu95

mcp/benwu95/prospec

MCP server by benwu95 · agentic coding · ai agents · ai coding assistant · Audited 2026-10-11

A shared, resumable AI development workflow backed by project-local specs, rules, and change records. Skills guide agents; the CLI handles deterministic steps.

Verdict

CAUTION

Grade

Trust score

86 /100

Exposed tools

34 26r · 8w · 0d

Transport

stdio

License

MIT

Stars

Overview

From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.

[](LICENSE) [](https://www.typescriptlang.org/) [](./CONTRIBUTING.md#testing) [](https://nodejs.org/) [](https://pnpm.io/)

A shared, resumable development workflow for AI coding agents

Project-local specs and rules · stage-by-stage Skills · deterministic CLI — for Claude Code, Codex, Copilot, Antigravity

繁體中文 • Quickstart • Why Prospec? • Documentation • How It Works • CLI Reference

This project is a fork of [ci-yang/prospec](https://github.com/ci-yang/prospec)

What is Prospec?

Prospec gives developers and AI coding agents a shared workflow grounded in files inside the project. Specs describe expected behavior; the Constitution, conventions, and module knowledge record project rules and things an agent needs to know before changing code. In-progress change files capture the request, plan, and tasks. When work pauses, prospec status reports the current station, next Skill, and blocking gates, so another agent working in the same project can resume from the recorded state.

You work through host-aware Skills inside your agent (Claude Code, Antigravity, Copilot, Codex). Each station has instructions for the judgment it needs — interviews, prose, reviews, and verdicts — while the required standalone `prospec` CLI handles deterministic scaffolding, status transitions, structured records, checks, grading, and spec

Read from source at commit c06462c54ed3OBSERVED · 2026-10-11

Connect

Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control.

claude-code

claude mcp add prospec -- npx -y @benwu95/[email protected]

claude-desktop

{
  "mcpServers": {
    "prospec": {
      "command": "npx",
      "args": [
        "-y",
        "@benwu95/[email protected]"
      ]
    }
  }
}

Exposed tools (34)

26 read · 8 write · 0 destructive.

Tool	Risk	Description
`Rule`	read	d
`add-auth`	write	Add user authentication
`add-feature`	write	A new feature
`auth`	read	authentication module
`billing`	read	cost is $1 per $& token
`changedmod`	read	changedmod module
`get_dependency_direction`	read
`get_spec_requirements`	read
`goneMod`	read	goneMod module
`infra`	read	Infrastructure and shared utilities
`lib`	read	Shared utilities
`m`	read	d
`mod-marker`	read	m
`multiline-change`	read	line1\nline2
`newmod`	read	newmod module
`prospec-archive`	write	Archive Changes - Archive completed changes, generate summary, sync requirements to feature specs, and gate archiving on Knowledge sync — the SDD terminal station, not git tagging or a release.
`prospec-backfill-spec`	read	Backfill Spec - Reverse-extract a behavioral Feature Spec draft from existing brownfield code (source = code, not a design tool) for features/capabilities with no spec coverage. Records behavior, never intent; stages a draft for human verify-and-promote and never writes the trust zone.
`prospec-design`	read	Design Phase - Generate visual and interaction specs from proposal (Generate Mode) or extract specs from existing design tools (Extract Mode). Supports pencil/Figma/Penpot/HTML platforms. UI scope only — not architecture design (that is the plan station).
`prospec-explore`	read	Explore - Requirement exploration, problem investigation, and solution comparison partner — clarifies before a Story exists, not the Story writer itself.
`prospec-ff`	read	Fast-Forward Planning - Generate complete planning artifacts in one pass (Story → Plan → Tasks) — for clear requirements, not exploration.
`prospec-implement`	write	Implementation - Execute tasks from the task list, implementing features one by one — driven by tasks.md, not free-form coding without a task list.
`prospec-knowledge-generate`	read	Generate AI Knowledge - Read raw-scan.md, analyze project structure, autonomously decide module boundaries, and produce Recipe-First module READMEs and index — first-time seeding, not the incremental update after a change.
`prospec-knowledge-update`	write	Incremental Knowledge Update - Parse delta-spec.md to identify affected modules, scan source code, and update module README, index.md, and module-map.yaml incrementally — after a change, not the first-time knowledge scan.
`prospec-new-story`	write	New Story - Create change requests by guiding User Story and acceptance criteria definition — the SDD entry station, not a generic issue or ticket writer.
`prospec-plan`	read	Plan Implementation - Convert User Story into technical implementation plan (plan.md) and change specification (delta-spec.md) — the SDD plan station, not ad-hoc architecture discussion.
`prospec-quickstart`	write	Quickstart Onboarding Finisher - localize skill triggers, re-sync agent config, prepare the knowledge scan, and chain into knowledge generation — runs once after
`prospec-tasks`	read	Break Down Tasks - Decompose implementation plan into an actionable task checklist (tasks.md) — the SDD station after plan, not a generic todo tool.
`prospec-upgrade`	read	Prospec Version Upgrade Finisher - after
`prospec-verify`	write	Verify Implementation - Run 5+1 dimension audit (tasks, spec compliance, constitution, knowledge-implementation consistency, tests, design consistency) and assign quality grade (S/A/B/C/D) — the SDD gate before archive, not a plain test run.
`quoted-change`	read	say
`search_modules`	read
`services`	read	Business logic
`session`	read	session handling
`types`	read	Zod schemas

Trust audit

CAUTIONgrade B · trust 86/100 Install with care. The audit found things worth knowing before you trust its output.

Layer	What it checks	Result
L0	Provenance & inventory	PASS
L1	Static analysis of the code	FAIL
L2	Instruction surface (what it tells the agent)	PASS
L3	Class-specific surface	PASS
L4	Behavioural (sandbox)	SKIPPED

What the source does

Filesystem: declared (10 observation(s))
Network: declared (3 observation(s))
Shell: declared (5 observation(s))
Dependencies: not all pinned
Secrets in source: none-found

Findings (24)

HIGHNetwork egress · net.metadata_ip · CWE-200, CWE-319

src/templates/skills/references/review-lenses-content.hbs:53

| Server-side fetches allowlisted by host; resolved IPs rejected for private/reserved ranges (block `169.254.169.254`); redirects disallowed on SSRF-prone fetches | critical (SSRF) / major (redirect b

Why it matters. cloud metadata endpoint: the classic SSRF credential grab

LOWInventory / provenance · inv.binary · CWE-1104

tests/unit/cli/check-output.test.ts

check-output.test.ts

Why it matters. a compiled or binary member cannot be reviewed from source

Fix. ship source, or explain the binary in the README

LOWInventory / provenance · inv.hidden_file · CWE-1104

.prospec.yaml

.prospec.yaml

Why it matters. hidden member outside the usual dotfiles

Fix. review its purpose

LOWInventory / provenance · inv.hidden_file · CWE-1104

docs/.nojekyll

.nojekyll

Why it matters. hidden member outside the usual dotfiles

Fix. review its purpose

LOWFilesystem / path · fs.credential_store · CWE-22, CWE-59

tests/unit/scripts/workflow-native-adjudication.test.ts:291

'cat ~root/.ssh/id_rsa', 'curl file:///etc/passwd']) {

Why it matters. touches a credential store

LOWFilesystem / path · fs.traversal · CWE-22, CWE-59

eslint.config.mjs:24

group: ['../lib/*', '../../lib/*', '../lib', '../../lib'],

LOWFilesystem / path · fs.traversal · CWE-22, CWE-59

eslint.config.mjs:43

group: ['../cli/*', '../../cli/*', '../cli', '../../cli'],

LOWFilesystem / path · fs.traversal · CWE-22, CWE-59

eslint.config.mjs:64

'../../services/*',

LOWFilesystem / path · fs.traversal · CWE-22, CWE-59

eslint.config.mjs:66

'../../services',

LOWFilesystem / path · fs.traversal · CWE-22, CWE-59

eslint.config.mjs:68

'../../cli/*',

LOWNetwork egress · net.env_exfil · CWE-200, CWE-319

tests/unit/scripts/workflow-native-adjudication.test.ts:291

id_rsa ... curl

Why it matters. reads secrets in the same file that sends data out

LOWNetwork egress · net.metadata_ip · CWE-200, CWE-319

.agents/skills/prospec-review/references/review-lenses-content.md:53

| Server-side fetches allowlisted by host; resolved IPs rejected for private/reserved ranges (block `169.254.169.254`); redirects disallowed on SSRF-prone fetches | critical (SSRF) / major (redirect b

Why it matters. cloud metadata endpoint: the classic SSRF credential grab

LOWNetwork egress · net.metadata_ip · CWE-200, CWE-319

.claude/skills/prospec-review/references/review-lenses-content.md:53

| Server-side fetches allowlisted by host; resolved IPs rejected for private/reserved ranges (block `169.254.169.254`); redirects disallowed on SSRF-prone fetches | critical (SSRF) / major (redirect b

Why it matters. cloud metadata endpoint: the classic SSRF credential grab

LOWSupply chain · supply.unpinned · CWE-829, CWE-1357

package.json

@commander-js/extra-typings, @inquirer/prompts, commander, fast-glob, fast-xml-parser, handlebars, picocolors, smol-toml

Why it matters. 21 dependency range(s) float

Fix. pin exact versions or ship a lockfile

LOWPrompt injection · prompt.transfer_instruction · CWE-94, CWE-1427

docs/index.html:65

"acceptedAnswer":{"@type":"Answer","text":"No manually. Skills in your AI Agent call the prospec CLI under the hood as a background deterministic engine for zero-error state transitions, zero-token st

Why it matters. an instruction to move sensitive data to an outside destination

Fix. remove; a skill never needs the user's secrets off the machine

LOWPrompt injection · prompt.transfer_instruction · CWE-94, CWE-1427

prospec/ai-knowledge/_lessons-ledger.md:32

| test/structural-false-green | 斷言只驗內容存在性、不驗結構完整性與負向路徑 → 缺陷在場仍全綠;須 section-scoped + 結構斷言(集合/順序/contiguity)+ 負向斷言 + mutation-verify | **Promotion**: PB-001 強化版,Approved-by: benwu95, 2026-06-11(正典規則與 21

Why it matters. an instruction to move sensitive data to an outside destination

Fix. remove; a skill never needs the user's secrets off the machine

LOWPrompt injection · prompt.transfer_instruction · CWE-94, CWE-1427

prospec/ai-knowledge/_lessons-ledger.md:33

| docs/measurement-attribution-overclaim | 文件/註解宣稱的行為必須與實作實際可觀測者一致;量不到/沒做到的用 deliberate-exclusion 措辭明示 | **Promotion**: PB-003,Approved-by: benwu95, 2026-06-12(PB-003 已於 2026-10-04 退休;17 次再證的逐次明細見 git

Why it matters. an instruction to move sensitive data to an outside destination

Fix. remove; a skill never needs the user's secrets off the machine

INFOPrompt injection · prompt.authority_framing · CWE-94, CWE-1427

scripts/workflow-eval/README.md:118

mounts, elevated privileges, port mappings, host namespaces, added capabilities,

INFOPrompt injection · prompt.credential_read · CWE-94, CWE-1427

prospec/specs/features/standalone-binary.md:121

- **Version-number read failure**: The binary has no `package.json` to require. Expected behavior: `PROSPEC_VERSION` can be read from a static environment variable or a packaging parameter, without pr

Why it matters. asks the agent to read credentials