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/mohammad19974/bounded-relay/adaptive-sddnpx skills add mohammad19974/bounded-relay --skill adaptive-sddgit clone --depth 1 https://github.com/mohammad19974/bounded-relayWrote 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/mohammad19974/bounded-relay/adaptive-sdd)<a href="https://agentmods.dev/skills/mohammad19974/bounded-relay/adaptive-sdd"><img src="https://agentmods.dev/badge/skills/mohammad19974/bounded-relay/adaptive-sdd.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 5.1 | $0.00034 | $0.00729 |
| Opus 5 | $0.00017 | $0.00365 |
| Sonnet 5 | $0.00007 | $0.00146 |
| Haiku 4.5 | $0.00003 | $0.00073 |
Grade A, and why
boundedrelay-adaptive-sdd 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 5d 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 — 70 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Adaptive dual-agent SDD
Use the packaged workflow as the source of truth. Do not improvise a second governance system.
Invariants
- Claude Code is the coordinator and keeps the host-selected model. Never force a Claude model.
- Route with
codex_worker_sdd_route. Hard eligibility and versioned task fit decide first. Use the default 5,000-basis-point Codex share only for tasks still neutral after fit; it is not a quota or a token, price, quality, or elapsed-time promise. Preserve and verify the returned policy versions, decision stages, and plan fingerprint. - Freeze Claude's review before starting Codex. Never place frozen Claude
findings in the Codex prompt or
focusfield. - Strict plan, implementation, and convergence gates use
codex_worker_sdd_review; generic analysis is advisory and cannot satisfy them. Accept only a current strict seal andgate.passed: true,gate.status: ready. - Execute the routed dependency graph in canonical waves. Start every wave from its verified clean committed checkpoint, finish all prior dependencies, and allow at most one writer. Codex write slices are revision-pinned isolated proposals with exact write scopes; BoundedRelay returns patches and never applies them. The coordinator reviews and integrates only within the user's existing authority.
- Keep one writer per repository at a time. Never provider-batch across waves, recursively delegate, silently downgrade a critical profile, auto-commit, push, publish, or deploy.
- Re-seal and repeat both reviews after any reviewed artifact, revision, policy, or implementation change. Missing, stale, malformed, or unavailable evidence fails closed.
- Complete every routed assignment exactly once. Zero Codex implementation assignments are valid when hard eligibility and task fit put all work on the host; record the host results without a worker call. Codex still participates in the independent review gates.
- Human rejection aborts the evidence chain. Start a fresh corrected run rather than re-prompting a stale gate or entering an unbounded provider loop.
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.
- 5d ago First seen · 70 lines · 34 tokens per session scan A b4cc4193a2d0
boundedrelay-adaptive-sdd is a skill published in the GitHub repository mohammad19974/bounded-relay (1 stars, last pushed 4d ago), licensed MIT. It adds 34 tokens to every session and 729 once invoked, about $0.0002 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
pipeline
Use when the user wants a feature idea taken end-to-end in one autonomous run — phrases like "run the whole pipeline", "take this feature from idea to finished branch", "brainstorm then build it autonomously", "do everything from idea to merged", "implement all phases without stopping". Triggers when they want…
craft
Use when a product idea is still vague and needs to become a clear definition of what to build — "let's craft an app like X", "help me define what I actually want", "clarify this idea before we plan it". Also use before planning or implementation when requirements, UX, domain behaviour, or technical preferences have…
deliver-cv-project
Turn an independently evaluated CV model or pipeline into a verified delivery handoff without copying volatile platform recipes. TRIGGER when: user has a working model, Workflow, OCR pipeline, or tracker and asks to export it, run locally, integrate an app/camera/RTSP source, deploy, productionize, monitor drift, add…
decision-report
Generate a portable stakeholder decision report for a CV deployment decision. TRIGGER when: user explicitly asks for a decision report, manager report, or stakeholder doc after a cost crossover has been computed; phrases: "write me a decision report", "generate a report for my manager", "I need a stakeholder doc"…
classify-or-flag
Produce one whole-image verdict: pass/fail, category, compliance, or anomaly. TRIGGER when: user asks whether an image is defective/compliant; to flag bad parts, classify product images, run image-level quality control, detect an anomaly, or build binary/multi-class classification. SKIP when: user asks how many…
detect-and-analyze
Detect instances and analyze boxes, counts, crops, regions, or metadata. TRIGGER when: user asks to count/detect objects, report the number of defects, build a model that counts trucks, tally boxes/pallets on a conveyor, find workers lacking PPE, crop detections, extract an ROI, estimate box-relative size, or report…