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 commands/rjmurillo/ai-agents/specgit clone --depth 1 https://github.com/rjmurillo/ai-agentsWrote 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/commands/rjmurillo/ai-agents/spec)<a href="https://agentmods.dev/commands/rjmurillo/ai-agents/spec"><img src="https://agentmods.dev/badge/commands/rjmurillo/ai-agents/spec.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 | $0.00016 | $0.04165 |
| Opus 5 | $0.00008 | $0.02083 |
| Sonnet 5 | $0.00003 | $0.00833 |
| Haiku 4.5 | $0.00002 | $0.00417 |
Grade A, and why
spec scanned grade A with 1 finding 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 today.
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.
Runs shell commandslowCapability
Expected in a hook, worth knowing in a rule or an instructions file.
**Invocation contract (security)**: the canonical flow is `Skill(skill="memory")`, which already passes topics via argv-vector internally. If the agent's environment lacks the `Skill` tool and must invoke the script dire How it starts
The opening of the file, as written. The whole thing — 177 lines — stays where its author put it; the contents beside it link to each section on GitHub.
@CLAUDE.md
Spec: $ARGUMENTS
If $ARGUMENTS is empty, ask the user what problem to solve. Do not proceed without a problem statement.
Process
Step 0: First Principles Gate (blocking, runs before Step 1)
Before any clarification work, answer six forcing questions. The gate exists because every retro citing wasted spec work in the last six months traces to a question this gate forces upfront. The strongest single citation is .agents/retrospective/2026-05-05-pr-1887-iteration-paradox.md Phase 6, where the retro itself names the question this gate asks ("is the framework worth building at all if its design space misses the dominant failure modes?") and explicitly defers it as out of scope. That deferral landed after 69 commits.
The six questions, asked in order:
| Label | Question |
|---|---|
| Q1 Demand Reality | Who has explicitly requested this? Name three or more individuals, teams, or systems by name. (Question is about requesters; production signals go to Q5.) |
| Q2 Status Quo | What is the exact workaround users do today, step by step? |
| Q3 Desperate Specificity | Name the single most blocked person or system right now. What exactly are they blocked on? |
| Q4 Narrowest Wedge | What is the smallest possible deliverable that unblocks Q3, measured in hours of implementation? |
| Q5 Observation | What direct production signal proves the gap exists? Cite a metric, log entry, error count, ticket, retro line, or trend. (Question is about signals; requesters go to Q1.) |
| Q6 Future-fit | If the system grows 10x, does this feature still make sense, or does it become a liability? |
Write the answers as a structured block (the ## Step 0 First Principles block) with six ### Q1..Q6 subheads, each containing the author's verbatim answer. The block flows downstream as input: Step 1 (Clarify) reads it as problem context, Step 2 (requirements-interview) carries it into the PRD it produces, Step 3 (Tier classification) re-validates Q4 at Tier 5, Step 6 (spec-generator) formalizes the PRD into durable artifacts with this block as the first section, and Step 9 (critic pre-mortem) checks that Q1/Q3/Q4 did not drift. Do not paraphrase; downstream steps depend on the verbatim answers.
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.
- today First seen · 177 lines · 16 tokens per session scan A f7da01531eeb
spec is a command published in the GitHub repository rjmurillo/ai-agents (45 stars, last pushed today), licensed MIT. It adds 16 tokens to every session and 4,165 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 1 finding (runs shell commands). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-03.
Other commands, from other repositories
eval
Manage eval-driven development workflow (define, check, report, list).
load-session
Load a synced session summary for context.
02_execute
Execute a plan (auto-moves pending to in-progress) with Ralph Loop TDD pattern.
update-model-rubric
Research the current LLM landscape and update Carmenta's model routing rubrics with latest models, capabilities, and pricing.
update-image-rubric
Research the current image generation landscape and update Carmenta's image model routing rubrics with latest models, capabilities, and pricing.
documentation-audit
Audit documentation for consistency across docs/, knowledge/ specs, and feature-catalog.ts - find gaps, misleading claims, and sync issues.