Claude Code Skills packaging nlink-jp workflows — development process, research, meeting minutes, security analysis, and news collection. This is an umbrella repository: each skill lives in its own repository and is included here as a submodule (ADR-004).
| Skill | Command | Description |
|---|---|---|
| rfp | /rfp [tool-name] |
Facilitates the RFP process for new nlink-jp projects. Collects requirements through Q&A, validates against CONVENTIONS.md planning phase, and outputs a structured RFP document. |
| mcp-tactics | /mcp-tactics |
Cross-cutting tactics book for the organization's MCP servers and proxies — decision tables from input artifact to route, cross-server chains, and a four-tier escalation doctrine ranked by who can observe the query (ADR-003, amended by ADR-018). |
| meeting-notes | /meeting-notes <transcript> |
Structures a meeting transcript (TXT/VTT/SRT) into a validated 3-layer JSON record — verbatim utterances, decisions with rationale, summaries — and compiles Markdown or self-contained HTML minutes. Successor to the meeting-note CLI. |
| service-research | /service-research <name> |
Researches a product or service — overview, ToS, privacy, data security, AI-agent behavior — by reading the primary sources on the web, and emits a schema-validated JSON report with a three-tier risk rating plus compiled Markdown (ADR-0001). Successor to the product-research CLI. |
| incident-research | /incident-research <incident> |
Deep-dives one publicly reported security incident — breach, ransomware, leak, supply-chain, exploited vulnerability — by collecting and reading news and primary sources, and emits a schema-validated JSON report (timeline-centric, three source tiers, confidence-qualified attribution), compiled Markdown, and a STIX 2.1 bundle of extracted IoCs (ADR-0001, ADR-0002). |
| incident-review | /incident-review <record> |
Retrospectively analyzes your own organization's IR communication record — Slack exports, plain-text logs, connector-read channels, or any transcript — behind a defang/nonce-isolation preprocessing gate, and emits a schema-validated JSON report (summary, activity, roles, process review) plus reusable tactic knowledge documents (ADR-0001). Successor to the ai-ir / ai-ir2 CLIs. |
| compliance-review | /compliance-review compile | review |
Two-phase security review against your own regulations — compile decomposes regulation documents into a versioned domain-expert set (full clause text per expert, no RAG, deterministic coverage check), review evaluates an application behind a sha256 drift gate and a nonce-isolation preprocessing gate with parallel independent experts, deterministic severity rules, and a two-pass Q&A (ADR-0001). Successor to the virtual-reviewer PoC. |
| news-digest | /news-digest --repo <corpus> |
Collects your own RSS/Atom/JSON feeds and digests only what is worth reading — the agent scores novelty, significance and relevance behind a nonce-isolation gate and never writes a priority, which a swappable profile's decision table derives. Continuing stories are tracked so a rehash with no new facts is named as such. The engine is public and holds no data; the feed list, relevance profile and article corpus live in a separate private corpus repository joined by a versioned contract (ADR-0001). First-time setup — corpus creation through scheduled operation — is a runbook the agent executes with the user. |
Each skill releases a zip whose root is the skill folder. Download it from the skill's own Releases page, then either:
unzip <skill>-vX.Y.Z.zip -d ~/.claude/skills/or upload the zip as-is to claude.ai / Claude Desktop under Settings → Skills.
To install from source instead, clone the skill's repository (or this
umbrella with --recurse-submodules) and run make install inside the
skill repository.
All skills follow the shared organization rules in
CONVENTIONS.md:
one repository per skill, skill content in a <skill-name>/ subdirectory,
structural validation via make check, and GitHub Release zips as the
distribution channel.
Each skill repository carries its own bilingual READMEs, CHANGELOG, and MIT license; this umbrella only tracks submodule pointers and the catalog above. The pre-split monorepo history (through v0.3.1) remains in this repository's git log and Releases.