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 joelbrilliant/agentic-delivery-skills --skill bulletproof-building-and-reviewgit clone --depth 1 https://github.com/joelbrilliant/agentic-delivery-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/joelbrilliant/agentic-delivery-skills/bulletproof-building-and-review)<a href="https://agentmods.dev/skills/joelbrilliant/agentic-delivery-skills/bulletproof-building-and-review"><img src="https://agentmods.dev/badge/skills/joelbrilliant/agentic-delivery-skills/bulletproof-building-and-review/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/joelbrilliant/agentic-delivery-skills/bulletproof-building-and-review"><img src="https://agentmods.dev/badge/skills/joelbrilliant/agentic-delivery-skills/bulletproof-building-and-review.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 | $0.01266 |
| Opus 5 | $0.00045 | $0.00633 |
| Sonnet 5 | $0.00018 | $0.00253 |
| Haiku 4.5 | $0.00009 | $0.00127 |
Grade A, and why
bulletproof-building-and-review 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 12d 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 — 118 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Bulletproof Building and Review
Overview
When building: prove the feature reaches production-shaped data, not just constructed test inputs.
When reviewing: prove that "tests pass" actually catches regressions.
Both sides share one discipline: verify against what production can actually produce, not what the test fixture constructs.
Companion Skill
Use orchestrated-agentic-delivery for the role workflow: orchestrator, builder, reviewer, reviewer-fix mode, and final signoff. This skill remains the evidence and verdict discipline: production-shaped data paths, test mapping, severity calls, and PASS/HOLD/REQUEST CHANGES.
The Iron Rule
Verify production-shaped behaviour, not math primitives. Trace upstream data paths. Find the writer for every field the reader consumes. If the writer does not exist, the feature is structurally inert. That is HIGH severity even if all tests pass.
Example: a scoring path read evidence.signal_id for empirical correlation weighting. The math was correct and all tests passed. But no ingestion path ever wrote signal_id, so the feature was permanently inert in production. The correct verdict was HOLD with HIGH severity: fix the ingestion contract.
When to Use
- Adversarial code review where you will write a verdict.
- Building a slice you will self-sign-off for peer audit.
- Reviewing scoring, security, ingestion, permissions, money movement, AI validation, or other production-bound logic.
- Before closing a finding as "not an issue".
- When the author claims pytest green, "tests exist elsewhere", or "I evaluated X and chose Y".
Do not use this for throwaway prototypes or work explicitly scoped as exploratory.
The 12 Rules
Semantic verification
Rule 1: Trace data flow upstream before signing off. For every field a function reads, locate the writer. No writer means structurally inert feature, which is HIGH severity.
Verdict structure
Rule 2: No "Blind Spots" section. Every uncertainty is either a Finding with severity and evidence, or Closed with evidence. No third bucket.
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.
- 12d ago First seen · 118 lines · 90 tokens per session scan A e5a626378ae4
bulletproof-building-and-review is a skill published in the GitHub repository joelbrilliant/agentic-delivery-skills (2 stars, last pushed 3mo ago), licensed MIT. It adds 90 tokens to every session and 1,266 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 2026-08-31.
Other skills, from other repositories
mcore-linting-and-formatting
Linting and formatting for Megatron-LM. Covers running autoformat.sh, tools (ruff, black, isort, pylint, mypy), and code style rules.
package-evaluator
Evaluates Claude Code package quality across 6 dimensions for all 7 package types, producing scored audit reports. Triggers on: "evaluate package", "audit agent quality", "score this hook", "package audit", "skill quality check". NOT for LLM prompts, use prompt-lab.
code-refiner
Deep code simplification and refactoring preserving behavior across Python, Go, TypeScript, Rust. Targets complexity, anti-patterns, readability debt. Triggers on: "simplify this code", "refactor for clarity", "reduce complexity", "make this more readable", "tech debt cleanup", "too much nesting".
codebase-advisor
Senior codebase advisor that audits a repository, vets findings, and writes self-contained implementation plans for other agents to execute. Triggers on: "write improvement plans", "create an implementation backlog", "audit and plan fixes", "turn findings into plans", "agent-executable plans", "reconcile plans"…
pr-review
Diff-based PR review across code quality, test coverage, silent failures, type design, and comment quality with severity-ranked findings. Triggers on: "review my PR", "review this code", "check my changes", "audit this PR", "code review". NOT for pre-landing gate, use pre-landing-review.
devils-advocate
Challenges AI-generated plans, code, and designs via pre-mortem, inversion, and Socratic questioning to surface blind spots and failure modes. Triggers on: "challenge this", "devils advocate", "stress test this plan", "poke holes in this", "what am I missing".