Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add stark-ai-de/agent-skills --skill architecture-compass-adr-authoringgit clone --depth 1 https://github.com/stark-ai-de/agent-skillsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/stark-ai-de/agent-skills/architecture-compass-adr-authoring)<a href="https://agentmods.dev/skills/stark-ai-de/agent-skills/architecture-compass-adr-authoring"><img src="https://agentmods.dev/badge/skills/stark-ai-de/agent-skills/architecture-compass-adr-authoring/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/stark-ai-de/agent-skills/architecture-compass-adr-authoring"><img src="https://agentmods.dev/badge/skills/stark-ai-de/agent-skills/architecture-compass-adr-authoring.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00090 | $0.01956 |
| Opus 5 | $0.00045 | $0.00978 |
| Sonnet 5 | $0.00018 | $0.00391 |
| Haiku 4.5 | $0.00009 | $0.00196 |
Grade A, and why
architecture-compass-adr-authoring scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 101 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Architecture Compass ADR Authoring
Goal
Turn rough decision notes into a reviewable first pull request that adds a reusable Architecture Compass decision and adopts the same decision in stark-ai-de/agent-skills, without making the maintainer restate the repository-specific PR prompt.
When to use
- A maintainer wants to add a new public, target-repository Architecture Compass ADR or propose a successor to one.
- The input is a rough idea, discussion, or partial requirements list.
- The requested outcome includes a pull request in
stark-ai-de/agent-skills.
When not to use
- The task is to adopt an existing Architecture Compass ADR in another repository; use
architecture-compass. - The decision is Architecture Compass implementation policy or belongs in
AC-INTERNAL-*; this paired-adoption workflow does not apply. - The task only changes implementation under an already accepted ADR.
- The user has not authorized repository writes or pull-request creation; stop after the requirements draft.
Inputs to inspect
- The user's decision intent, constraints, examples, and explicit exclusions.
- Current
main, repository instructions,docs/adrs.md,docs/adrs/TEMPLATE.*.md, and relevant repository ADR triplets. skills/engineering-workflows/architecture-compass/SKILL.md,skills/engineering-workflows/architecture-compass/references/adr-catalog.md, and relevant exposed ADR triplets.scripts/validation/architecture-compass/validate.mjs, its decision lock and lineage files, release contracts, plugin source configuration, and generated projection rules.- Open issues and pull requests that may already cover the same decision.
Workflow
- Resolve authority, isolation, and baseline. Confirm the target repository and requested PR outcome from the user's request. Require one assigned external worktree for every repository write, keep the canonical checkout read-only, and work from current
main. If that isolation cannot be established, stop with concrete setup guidance. Never hand-edit generated plugin projections. - Draft generic requirements first. Convert the source request into a compact bullet list. Keep one self-contained, independently reviewable requirement per bullet, make the wording reusable across projects, separate outcomes from implementation notes, preserve explicit constraints, and include validation or failure behavior when material. Remove repository paths, personal names, and one-off examples unless they are essential to the decision. Record assumptions separately instead of silently converting them into requirements.
- Check overlap and conflicts. Search both ADR namespaces by decision, trigger, tags, and affected boundary. Classify the proposal as new, overlapping, conflicting, or a successor. Do not rewrite an accepted decision in place. Continue preparing the PR when conflicts exist, but identify every relevant ADR and the unresolved impact.
- Resolve namespace, status, and identities. This workflow accepts only an exposed
AC-ADR-NNNwithScope: target-repository; reclassify an internal or skill-runtime decision and stop before paired authoring. The shipped Architecture Compass inventory permits onlyAcceptedorSuperseded. For a new record, require the maintainer to approve the intended post-mergeAcceptedstatus before authoring; without it, return the requirements and conflict draft and stop. Default only the repository-local adoption toProposedunless its status is separately approved. Then select independent unused IDs from the live catalogs. - Author the paired triplets. Add Short, Long, and Guide variants under the Architecture Compass references and under
docs/adrs/. Derive the provider triplet from current exposed siblings and the validator; Architecture Compass's bundledassets/adr-template.*.mdfiles are target-repository templates, not provider-authoring templates. Use the repository's own templates or generator for the local triplet. Long is canonical and contains one decision. Short is a faithful, compact, on-point abstraction with only the essential context, outcome, and trade-offs. Guide contains non-normative application, examples, verification, current sources, and required lineage. - Synchronize owning surfaces and component versions. Update the complete approved Architecture Compass validator inventory, catalog, lineage manifest, accepted-decision lock, repository ADR index, and any required routing or eval surfaces. Update repository instructions only where an accepted local decision must become effective. Bump the promoted skill version and affected plugin version, and update derived listing surfaces required by the live feature-impact contract. Reserve the root package version, release manifest, and changelog for a separate generated Release Please PR under ADR-0050. Run the repository generator for the portable plugin projection; do not edit
plugins/stark-ai-developer/directly. - Validate changed contracts. Run the focused ADR, skill, projection, release-intent, formatting, and diff checks described in the reference, plus owners for any changed release or listing surfaces. Run
npm run list:incubatoronly if incubator discovery changed. Distinguish component impact from root release intent. Run the local aggregate after focused checks stabilize only when ADR-0041 or another mandatory gate requires it; the hosted PR aggregate remains mandatory. - Create the first PR. Use a focused branch and conventional commit title. Put the normalized requirements first in the PR body, followed by the paired ADR identities and paths, conflict report with decision owners, synchronization performed, candidate-bound validation evidence, unavailable checks, and all remaining decisions. Once the status gate is satisfied, a conflict is review information and is not a reason to omit either proposed triplet from the PR.
What ships with it
2 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 3d ago First seen · 101 lines · 90 tokens per session scan A dbe61adb3030
architecture-compass-adr-authoring is a skill published in the GitHub repository stark-ai-de/agent-skills (5 stars, last pushed yesterday), licensed Apache-2.0. It adds 90 tokens to every session and 1,956 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-09.
Other skills, from other repositories
levelup-publish
Compile accepted Context Directive Records (CDRs) into team-ai-directives artifacts and create a draft PR. Builds context modules, evals goldensets, and/or skills based on CDR context types.
change-init
Mine git history for Change Decision Records (ChDRs) by detecting commit messages that link to issue trackers, clustering the commits into change stories, and inferring the decisions behind them. Use when bootstrapping project memory from an existing repo's history (brownfield), before refactoring unfamiliar code, or…
workspace
Multi-repo workspace coordinator for shared team context. Initialize .adlc/ structure, configure .gitignore, discover child repos, link them as Git submodules, and audit workspace health. Use --init for first-time setup, default mode for ongoing auditing.
git-under-contention
Operate Git safely when multiple humans or coding agents may share a repository, checkout, index, branches, or worktrees. Use before staging, committing, pushing, merging, rebasing, creating or removing worktrees, recovering apparently lost edits, handling staged files you did not touch, or preserving unfinished agent…
repo-standard
Use when working in a RepoKit-compliant repo — about to commit or open a PR, adding a learning / ADR, or asking "where does X go" or "what's the convention here". Loads the RepoKit conventions and the pre-commit / pre-PR checklists.
cp
Local verify-commit-push wrapper for Codex. Use only when the user explicitly says /cp or clearly asks to finish the current work by running the relevant local checks, staging intended files explicitly, committing on the current branch, and pushing. This workflow may commit directly to main or master when that is the…