mantis-architecture
此技能把零散洞察与代码库结构分析合成为一份规范化、互相链接的 Markdown 知识库:定义系统架构、梳理实体组件、归类历史漏洞模式,并把结果带快照标记地持久化到 workspace/kb。它让后续智能体复用同一份架构记忆,保证知识的可追溯性与新鲜度。
npx skills add https://github.com/google/mantis --skill mantis-architectureBefore / After 效果对比
2 组开发者每次接手陌生代码库都要手动翻阅目录结构、追踪导入关系并口头讲解组件职责,架构上下文散落在个人笔记里,一次完整梳理往往耗掉大半天。
该 Skill 自动产出 architecture.md、entities 与 vulnerabilities 条目及 dependencies.json,组件职责与依赖边一目了然,后续智能体直接读取即可开工。
Architect (/mantis-architecture)
System Goal
Knowledge Base Synthesizer. Translates ephemeral insights from the learnings
queue (workspace/learnings.jsonl) and structural analysis of the codebase into
a canonical, interlinked Markdown Knowledge Base (workspace/kb/).
Command Definition
- Command:
/mantis-architecture - Description: Builds the foundation of the KB by defining system architecture, mapping specific entities (components), and categorizing historical vulnerability patterns.
- Arguments (all optional; resolved by LOCATOR RESOLUTION / Block A):
--snapshot_root=<dir>(or theSNAPSHOT_ROOTenv var) — the pinned code snapshot to read target source from;--snapshot_id=<id>— theSNAPSHOT_IDof that snapshot;--state_root=<dir>— the parent ofworkspace/for all state and KB paths;--target_root=<dir>— an already-prepared tree that OVERRIDES the snapshot (rarely passed to this stage). When none are passed, behavior is byte-for-byte today's (degraded/unpinned): read source from the current directory and treatsnapshot_pinnedas false.
Input/Output Contract
- Reads:
workspace/learnings.jsonl(raw insights from the current round).workspace/historical_learnings.jsonl(optional, past vulnerability metadata).- Codebase directory structure and key source files.
- Existing Markdown files in
workspace/kb/(to validate/decay check). workspace/.mantis_state.json(to retrieve pass count).workspace/.mantis_state.json→active_snapshot(root,snapshot_id,snapshot_pinned) andsnapshot_history— provenance for the KB freshness gate (step 0b). Read the snapshot from STATE ONLY; NEVER run a live VCS command (git/hg/repo) to decide KB currency.workspace/.mantis_state.json→kb_snapshot_id(theSNAPSHOT_IDthe current KB was last built against; absent on a first/legacy KB).workspace/.mantis_state.json→changed_filesandchanged_files_status(written by mantis-plan's Block E; consumed by the scoped KB invalidation in step 0b outcome 3. Present from pass 2 on, but may be stale (written in a prior pass) — the scoped path checkschanged_files_passagainststate.pass_numberand falls back to full rebuild if they differ.)
- Writes:
- Markdown files under
workspace/kb/(architecture.md,entities/[component_name].md,vulnerabilities/[CWE-ID].md,index.md). workspace/kb/dependencies.json— a JSON map of import/dependency edges extracted during architectural analysis (keys = source file paths relative to CODE_ROOT; values = arrays of files that import/depend on the key file). This is consumed bymantis-plan's dependency-aware fan-out (Phase 2). If the codebase has no parseable import structure, write{}. Re-derive only changed entries during scoped invalidation (see outcome 3 below).- Archives
workspace/learnings.jsonltoworkspace/archive/learnings/learnings_pass_${N}_${X}.jsonl. - A
<!-- KB_SNAPSHOT: <SNAPSHOT_ID> -->marker as the FIRST line of every (re)written KB file (index.md, eachentities/*.md, eachvulnerabilities/*.md), pluskb_snapshot_id=SNAPSHOT_IDinworkspace/.mantis_state.json. - An immutable per-pass copy of the whole KB tree to
workspace/archive/kb/kb_pass_${N}_${X}/(so a later reverted fix cannot silently erase the record of what the KB claimed at pass N).
- Markdown files under
- Preconditions:
workspace/learnings.jsonlmust exist.
- Idempotency Guarantee:
- Transactional: moves
workspace/learnings.jsonlto archive only after programmatically verifying all KB Markdown updates were written successfully. KB files are overwritten in-place. - Snapshot stamping is part of the same transaction: the
KB_SNAPSHOTmarkers, the per-passworkspace/archive/kb/kb_pass_${N}_${X_kb}/copy, and thekb_snapshot_idstate write complete before (or together with) the learnings move. On any failure, leaveworkspace/learnings.jsonlintact.
- Transactional: moves
Instructions
Analyze the codebase and pending learnings to construct a permanent, Markdown-based memory for future agents.
Execute the architecture stage as follows:
- LOCATOR RESOLUTION (Block A, inlined below):
LOCATOR RESOLUTION (before reading ANY target code or artifact):
0. ROLE: If this skill NEVER reads target source (report, calibrate, reflect),
you are a FINDINGS-ONLY stage: skip steps 2-6; still read active_snapshot from
state for provenance/annotation; NEVER stop merely because a code root is unset.
1. Determine CODE_ROOT, in this priority order:
a. If --target_root is passed on THIS invocation, CODE_ROOT = --target_root.
It is AUTHORITATIVE and OVERRIDES SNAPSHOT_ROOT and the state fallback
(used when a caller hands you a prepared tree, e.g. a patched shadow).
b. Else if --snapshot_root (or SNAPSHOT_ROOT) is passed, use it.
c. Else read state_root/workspace/.mantis_state.json (state_root from
--state_root if passed, else ./workspace/... relative to the current dir)
-> active_snapshot.root / .snapshot_id / .snapshot_pinned.
d. Else (no arg AND no readable active_snapshot): CODE_ROOT = current directory,
treat snapshot_pinned = false (MODE-OFF). Do NOT stop.
2. SENTINEL CHECK (only if snapshot_pinned is true AND you did NOT take path 1a):
verify CODE_ROOT/.mantis_snapshot_id exists and equals SNAPSHOT_ID. If missing
or different -> STOP "snapshot sentinel mismatch". (A --target_root tree (1a) is
deliberately mutated and is sentinel-EXEMPT.)
3. PATH FIELDS:
- SNAPSHOT-RELATIVE (read under CODE_ROOT): code_paths entries; plan target_files
that are file paths. Strip ONLY a trailing ":<digits>". A code_paths entry
containing "://" is a URL/endpoint, NOT a file read. A code_paths entry that is
NOT of the form <existing-path>:<integer> is a non-source LOCATOR
(symbol/offset/endpoint): only check that the artifact/symbol exists; skip ALL
line-range and line-existence logic.
- STATE-RELATIVE (read/write under state_root/workspace, NEVER prefix CODE_ROOT):
kb_references, repro_file_path, reattack_file_path, helper scripts, report
files, and all state/findings JSON.
4. Never WRITE under CODE_ROOT when snapshot_pinned is true. Any command that
compiles, generates, or writes artifacts MUST run in a PRIVATE SHADOW copy
(mktemp -d from CODE_ROOT), never with cwd=CODE_ROOT. Read-only inspection may
cd into CODE_ROOT.
5. VCS-METADATA CARVE-OUT: history-log extraction and any VCS diff/blame command
run in the LIVE repository root (which still has .git/.hg/.repo), NOT CODE_ROOT
(the snapshot copy strips VCS metadata). Do NOT stop merely because CODE_ROOT
lacks .git/.hg/.repo.
6. Every shell command uses ABSOLUTE paths and sets its own working directory on
that call. Do NOT assume the working directory persists between calls.
This is a CODE-READING stage (steps 2 and 4 read target source), so Block A
steps 1-6 all apply; it is NOT findings-only. Resolve CODE_ROOT,
SNAPSHOT_ID, and snapshot_pinned from state BEFORE doing anything below. Per
Block A step 3, all workspace/kb/... paths are STATE-RELATIVE: read and write
them under state_root/workspace, NEVER under CODE_ROOT. Read all target
source under CODE_ROOT. Do NOT run any VCS command to decide KB freshness
(Block A step 5's carve-out is only for history/diff/blame, which this stage
does not use).
0b. KB SNAPSHOT FRESHNESS GATE (mechanical; STATE + KB marker only, NO live VCS):
- `CUR` = `SNAPSHOT_ID` (resolved by Block A; the empty string if
`active_snapshot` was absent). `PINNED` = `snapshot_pinned` (false if absent).
- **MODE-OFF short-circuit (3-state rule):** if `active_snapshot` is ABSENT
in state (no `--sync` was requested — MODE-OFF = today's default), SKIP the
freshness gate entirely: do a best-effort build/update against `CODE_ROOT`,
do NOT prepend any STALE banner, and do
NOT stamp `kb_snapshot_id`. This is byte-for-byte today's behavior. (Only
HALT and PINNED run the gate below.)
- `KB_ID` = the `kb_snapshot_id` value in `.mantis_state.json` (primary); else the
text after `KB_SNAPSHOT:` on the FIRST line of
`state_root/workspace/kb/index.md` if that file exists (secondary fallback,
for legacy runs without state); else `""` (no prior KB). State is primary so
the file-marker parsing pitfall (comment-wrapped first line, no `-->`
stripping) can never strand `KB_ID` with the comment closer and force BUILD
FRESH every pass.
- Choose EXACTLY ONE outcome by string checks, top to bottom, first match wins:
1. `PINNED` is false (HALT mode — `active_snapshot` present but unpinned)
-> **STALE / HALT.** Do a best-effort build/update against `CODE_ROOT`,
but PREPEND the STALE banner (below) as the first lines of `index.md`.
Do NOT claim currency: leave the
banner in place. Set `kb_snapshot_id` = `CUR` (a `live:` id).
2. Else `KB_ID` == `CUR` (both non-empty) -> **CURRENT.** Do the incremental
update + decay-check (step 4) as today. Re-stamp `KB_SNAPSHOT: CUR` on
every (re)written file. REMOVE any STALE banner previously prepended to
`index.md`.
3. Else (`PINNED` true AND (`KB_ID` is empty OR `KB_ID` != `CUR`)) ->
**BUILD FRESH (full or scoped re-architecture).** The pinned code advanced
since the KB was built (a sync / pass-boundary change), OR the KB is
unstamped / legacy. Choose full or scoped:
- **Scoped invalidation (Phase 2 incremental efficiency):** If
`changed_files_status` is known (not UNKNOWN) AND the KB already has a
`KB_SNAPSHOT` stamp (KB_ID was non-empty, just different) AND
`changed_files_pass` equals the current `state.pass_number` (the diff
is from THIS pass, not a stale prior pass — absent or different →
treat as UNKNOWN → full rebuild below), attempt a SCOPED rebuild: only
invalidate KB entries whose source files are in `changed_files`, plus
their parent-rollup dependents (KB entities that import/reference the
changed files). Re-derive ONLY those entries from `CODE_ROOT`; carry
forward all other KB entries unchanged (they were built against the
same code, just a different snapshot ID). Re-stamp `KB_SNAPSHOT: CUR`
on every (re)written file.
- **Parent-rollup (2-hop, matching plan's fan-out):** When
invalidating a KB entry for changed file F, also invalidate any KB
entry that REFERENCES F directly (1-hop) AND any entry that
references a 1-hop dependent of F (2-hop). This matches
`mantis-plan`'s dependency-aware fan-out (which expands up to 2
hops), ensuring that a grandchild entity (H imports G, G imports
changed F) is not carried forward stale and later fed as a
`kb_reference` while its dependency has changed.
- **Guardrail:** If ANY uncertainty arises (can't determine which KB
entries map to which source files, the KB structure is ambiguous, or
changed_files is empty but KB_ID != CUR), fall back to full rebuild
below. Never carry forward a stale entry for a changed file.
- **Full rebuild (Phase-1 fallback):** If KB_ID is empty (no prior KB),
OR `changed_files_status` is UNKNOWN, OR the scoped invalidation guardrail
fired, REBUILD every KB file from scratch against `CODE_ROOT`. Do NOT
carry forward any prior assertion you have not re-derived from
`CODE_ROOT` this pass. Stamp `KB_SNAPSHOT: CUR`. REMOVE any STALE banner.
- STALE banner (paste verbatim, substituting `<KB_ID>` and `<CUR>`; keep the
leading `>` on every line so it renders as a visible blockquote):
```
> **STALE KB WARNING — do not trust without re-verifying.**
> snapshot_pinned=false, or the KB was built against a different snapshot.
> KB_SNAPSHOT=<KB_ID> does not match active_snapshot.snapshot_id=<CUR>.
> Every SECURE/FIXED/NON_VIABLE/SAMPLE_OR_TEST claim below is UNVERIFIED
> against the current code. Re-verify before trusting; do NOT filter, skip,
> or down-prioritize work based on this KB.
```
-
Read the Inbox (
workspace/learnings.jsonlandworkspace/historical_learnings.jsonl):- Parse the contents of
workspace/learnings.jsonl(andworkspace/historical_learnings.jsonlif it exists). Extract all trajectory insights, discovered vulnerabilities, viable crash paths, and verified patches.
- Parse the contents of
-
Analyze Source Code Boundaries:
- Examine the directory structure and key source files under
CODE_ROOT(the pinned snapshot resolved by Block A; use absolute paths per Block A step 6, and do NOT run a VCS command). Dynamically identify the core components, interfaces, and trust boundaries of the system based on the repository's contents. This applies broadly across domains: whether it is a software system (e.g., identifying parsers, controllers, or network daemons), a hardware/RTL design (e.g., identifying IP blocks, JTAG interfaces, or memory controllers), Infrastructure-as-Code (e.g., identifying cloud permissions, VPC perimeters, or deployment descriptors), or data/ML pipelines (e.g., identifying data ingress points, model serialization mechanisms, or training boundaries).
- Examine the directory structure and key source files under
-
Build or Update the Knowledge Base (KB):
-
Create or update files in the
workspace/kb/directory using standard Markdown. Follow these strict paths:workspace/kb/architecture.md: High-level data flows, zone definitions, system design, and overall availability/uptime requirements (if documented or inferable from configuration like systemd, kubernetes, or load balancers).workspace/kb/entities/[component_name].md: Specific definitions for components (e.g.,auth_module.md). Must include links to associated vulnerability classes and document known constraints (e.g., "This module sanitizes input X"). Document the component's criticality and availability requirements (classify as CRITICAL, STANDARD, or LOW_CRITICALITY if applicable). Incorporate trajectory insights here.workspace/kb/vulnerabilities/[CWE-ID_or_BugClass].md: Descriptions of bug classes (e.g.,CWE-79.mdorMemory-Corruption.md) that have been historically relevant to this codebase, including examples of what not to do.workspace/kb/index.md: A root catalog containing links and 1-line summaries to every file created above. This is the map the Planner will read.
-
Important Formatting Rules: Use relative links to cross-reference entities and vulnerabilities (e.g.,
[Auth Module](entities/auth_module.md)). Ensure all markdown files are concise and focused on actionable security context.
-
...
用户评价 (0)
发表评价
暂无评价
统计数据
用户评分
为此 Skill 评分