sdlc-evals

sdlc-evals is a command for coding agents from MCKRUZ/claude-code-sdlc. It costs 0 tokens per session (1,235 once invoked), scanned A, original, MIT.

An evaluation-authoring command creates a versioned set of example scenarios and expected outcomes for a specification that uses an AI-powered feature. These examples act as acceptance criteria for behavior that may vary between runs.

In plain words
What is it for?
Use it to prepare a golden set alongside an SDLC specification, working with business requirements and data definitions in either a project or standalone repository.
Why use it?
It turns vague expectations into agreed checks and helps teams decide what good responses or results should look like before implementation or testing.

Command

Part of the claude-code-sdlc plugin — 9 skills, 28 commands, 20 agents shipped together

Install

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.

agentmods
npx agentmods add commands/mckruz/claude-code-sdlc/sdlc-evals
Clone the repo
git clone --depth 1 https://github.com/MCKRUZ/claude-code-sdlc

Or install claude-code-sdlc, the plugin that ships this one along with the rest of its 9 skills, 28 commands, 20 agents.

Wrote 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.

agentmods badge for sdlc-evals

README.md
[![agentmods](https://agentmods.dev/badge/commands/mckruz/claude-code-sdlc/sdlc-evals.svg)](https://agentmods.dev/commands/mckruz/claude-code-sdlc/sdlc-evals)
Your own site
<a href="https://agentmods.dev/commands/mckruz/claude-code-sdlc/sdlc-evals"><img src="https://agentmods.dev/badge/commands/mckruz/claude-code-sdlc/sdlc-evals.svg" alt="Measured on agentmods" height="20"></a>
Per session 0 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,235 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00000 $0.01235
Opus 5 $0.00000 $0.00617
Sonnet 5 $0.00000 $0.00247
Haiku 4.5 $0.00000 $0.00123

Measured 3d ago against content hash bf527c4f8ac8, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

sdlc-evals 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.

commands/sdlc-evals.md · 77 lines

How it starts

The opening of the file, as written. The whole thing — 77 lines — stays where its author put it; the contents beside it link to each section on GitHub.

/sdlc-evals — Author the Versioned Golden Set for an LLM-Powered Spec

Author the golden set — the acceptance criteria for probabilistic behavior — for a spec on an llm_powered channel. Bizreq owns the scenarios, Data owns the data; the two are authored collaboratively. This command is a thin wrapper around the existing eval-builder harness skill: it adds no new script, authors no CI YAML, and writes a versioned golden-set.yaml next to the spec. Interview-driven like /sdlc-coach at the collaboration and calibration steps — it proposes; a named human decides. Works inside an SDLC project or standalone.

Instructions

  1. Resolve context:

    • Workflow mode (default): .sdlc/state.yaml exists. The repo root is the directory containing .sdlc/; read the target spec and its golden-scenarios.md (SCEN-NN) and data-contract.md.
    • Standalone mode (--repo <path>, or no .sdlc/ found): operate on the given repo with provisional context; take scenarios/data from the paths given or asked for.
  2. Confirm the channel is llm_powered: Read the target spec's channel: field and channels/<channel>.yaml. If llm_powered: true, evals are required acceptance criteria — the descriptor's eval_hooks name which dimensions become golden-set cases. If llm_powered is false or the spec has no channel, tell the human evals are advisory here and stop unless --force is given.

  3. Gather the inputs (collaborative): Assemble the cases from Bizreq's SCEN-NN golden scenarios (the behaviors that must hold) and Data's representative data (real inputs from the manual pre-release checks and the bug/support queue, not invented cases), plus the descriptor's eval_hooks. Ask focused questions to fill gaps — which failures matter, what the reference answer is for each.

  4. Wrap the eval-builder skill (no new script): Follow the eval-builder skill's procedure (harness/skills/eval-builder/SKILL.md, or the target repo's installed .claude/skills/eval-builder) — start from real failures small (20–50 tasks), pick graders deterministic-first (state_check / transcript_constraint over an llm_rubric, and grade the output, not the trajectory), compose multidimensional success where needed, and write the versioned golden-set.yaml next to the spec using the skill's shape. The skill owns the output path and the template — do not hardcode either here (it writes under eval-datasets/specs/<feature>/).

Read the full file on GitHub · 77 lines

Changes

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.

  1. 3d ago First seen · 77 lines · 0 tokens per session scan A bf527c4f8ac8

Subscribe to this mod's changes

sdlc-evals is a command published in the GitHub repository MCKRUZ/claude-code-sdlc (4 stars, last pushed 6d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 1,235 tokens. 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.