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 agentmods add skills/blakemartz/delivery-loop/author-styleguidesnpx skills add blakemartz/delivery-loop --skill author-styleguidesgit clone --depth 1 https://github.com/blakemartz/delivery-loopWrote 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/blakemartz/delivery-loop/author-styleguides)<a href="https://agentmods.dev/skills/blakemartz/delivery-loop/author-styleguides"><img src="https://agentmods.dev/badge/skills/blakemartz/delivery-loop/author-styleguides.svg" alt="Measured on agentmods" 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 | $0.00089 | $0.00986 |
| Opus 5 | $0.00044 | $0.00493 |
| Sonnet 5 | $0.00018 | $0.00197 |
| Haiku 4.5 | $0.00009 | $0.00099 |
Grade A, and why
author-styleguides 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 4d 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 — 38 lines — stays where its author put it; the contents beside it link to each section on GitHub.
author-styleguides — seed and evolve the constitution
You author the conventions the loop treats as constitutional: the decomposer fits tasks to them, implementers read them before writing code, and reviewers enforce them as the best-practice bar (spec ${CLAUDE_PLUGIN_ROOT}/docs/agentic_delivery_spec.md §2.1, §3.2–3.4). They are the prose counterpart of the gate: GATE_CMD defines correct mechanically; the styleguides define good where only judgment can.
Read .claude/delivery.conf for STYLEGUIDES_DIR (default styleguides/), GATE_CMD, and BASE_BRANCH.
Inputs
- Optional scope argument (a stack or layer, e.g. "python", "frontend") — default: every stack detected in the repo.
- What already exists in
STYLEGUIDES_DIR. This skill is re-runnable: add the missing guide when a stack arrives, or revise a ratified one on explicit request — never silently rewrite ratified text. - The repo itself: language/framework manifests, the gate script
GATE_CMDpoints at (it shows which rules are already enforced mechanically), CI workflows, and the existing code.
Protocol
- Detect. Enumerate the stacks and layers actually present (languages, frameworks, test tooling, repo topology — single package vs monorepo). One guide per stack/layer that agents will touch; skip layers with no code and no imminent work.
- Observe. For each stack with existing code, mine the dominant conventions: layout, naming, typing strictness, error handling, dependency management, test placement and idiom. Where the codebase is consistent, the guide codifies what is. Where it is inconsistent, that is a fork to pin, not a convention to codify. Greenfield stacks start from the stack's idiomatic best practice.
- Pin. Put the genuinely contested choices to the human as explicit decision sets, recommendation first (AskUserQuestion) — strictness levels, structural patterns, test philosophy, how to resolve each observed inconsistency. Never write a rule over an unpinned fork.
- Draft one guide per stack/layer in the scratchpad. Keep each readable in one sitting: prescriptive rules, a short right/wrong example where ambiguity is likely, no restating what the stack's defaults already guarantee.
- Checkable rules graduate to the gate. If a rule can be enforced mechanically (formatter, linter rule, type-checker flag, test), the guide states the principle and the enforcement belongs in the gate — wire it into
GATE_CMD's script now if trivial, otherwise flag it as a follow-up task for the backlog.
- Checkable rules graduate to the gate. If a rule can be enforced mechanically (formatter, linter rule, type-checker flag, test), the guide states the principle and the enforcement belongs in the gate — wire it into
- Ratify — hard gate. Present the drafts to the human with every judgment call flagged (each pin taken, anything observed-but-surprising, rules you chose not to write). Nothing lands in the repo before the nod. Apply their edits; re-present if the changes are substantive.
- Land. Write the ratified guides into
STYLEGUIDES_DIR(replace the seeded README stub's "empty constitution" note if present). Commit toBASE_BRANCHdirectly when the loop isn't live yet; open a PR when it is or when the operator prefers.
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.
- 4d ago First seen · 38 lines · 89 tokens per session scan A fd6c76ebac22
author-styleguides is a skill published in the GitHub repository blakemartz/delivery-loop (4 stars, last pushed 1mo ago), licensed MIT. It adds 89 tokens to every session and 986 once invoked, about $0.0004 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 2026-08-31.
Other skills, from other repositories
codex-thread-orchestration
Use when running Arcgentic V2 in Codex and the current thread must orchestrate fixed Planner, Developer, Test, and Auditor role threads.
claude-code-session-broker
Use when running Arcgentic V2 in Claude Code and fixed Planner, Developer, and Auditor role sessions must be coordinated through a broker.
arcgentic
Use when the user says Arcgentic, asks to use Arcgentic, or wants an idea taken through a complete plan → development → self-audit → external audit workflow in Codex.
audit-round
External-audit role for arcgentic rounds. Loaded when arcgentic state.yaml is in awaitingaudit or auditinprogress, OR when the user explicitly requests an external audit of a finished round, OR when reviewing a handoff doc + commit chain against the round's declared scope. Produces a PASS / NEEDSFIX / AUDITINCOMPLETE…
verify-gates
Runs the mechanical quality gates that the arcgentic state machine requires for state transitions. Invoked indirectly by transition.sh OR directly by orchestrator agent before declaring a state transition. Use when about to call transition.sh OR when manually verifying that a round artifact meets the gate criteria.…
session-mode
Use when a project has not yet stored session mode, when a user asks for complete arcgentic workflow execution, or when role identity handoff prompts are needed.