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/juanklagos/spec-driven-development-template/servegit clone --depth 1 https://github.com/juanklagos/spec-driven-development-templateWhat 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 | $0.00033 | $0.00507 |
| Opus 5 | $0.00016 | $0.00253 |
| Sonnet 5 | $0.00007 | $0.00101 |
| Haiku 4.5 | $0.00003 | $0.00051 |
Grade A, and why
serve 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 2d 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.
What it actually says
You are serving the SDD Builder's AI request queue (spec 031). The user is working on the canvas and pressing "Ampliar con IA" on fields; each press puts a request in the queue for you. Respond in the user's language (EN/ES).
Loop:
- Call
sdd_next_request(projectRoot: the current workspace, agent:claude-code). - If it returns
request: null, say so in one line and stop — or keep polling if the user asked you to stay listening (/loopworks well for this). - If it returns a request, read all of it:
targetsays which field it is (spec.md section, task, note, or logbook entry),currentTextis what exists today,instructionis what the person asked for. - Draft ONLY the proposed text for that field:
- same language as the current text, same format (a list returns one line per item);
- criteria keep the EARS pattern
WHEN … THE SYSTEM SHALL …/CUANDO … EL SISTEMA DEBERÁ …; - for
structure-idearequests, return exactly the JSON the instruction asks for, with no markdown fences around it.
- Answer with
sdd_respond_request(the request id,proposal: your text). - Repeat from step 1.
Hard rules:
- Do NOT write any file under
specs/, and do not usesdd_update_spec_sections,sdd_write_spec_document,sdd_add_taskor any other write tool while serving. Your proposal is not applied: the person reviews it as a diff in the builder and only their acceptance writes. - No inventes contexto que no esté en la petición o en el workspace. Si algo es ambiguo, propón la mejor versión y dilo en una frase al final.
- Do not ask for permission between requests: the permission IS the builder's Accept button.
If the queue tools are missing, the MCP server is not connected — tell the user to run npx @juanklagos/sdd-mcp@latest connect and restart this client. / Si faltan las tools de la cola, el MCP no está conectado.
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.
- 2d ago First seen · 26 lines · 33 tokens per session scan A 6cb287e4ef71
serve is a command published in the GitHub repository juanklagos/spec-driven-development-template (15 stars, last pushed 2d ago), licensed MIT. It adds 33 tokens to every session and 507 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 2026-08-30.
Other commands, from other repositories
paul:help
Show available PAUL commands and usage guide.
paul:verify
Guide manual user acceptance testing of recently built features.
auto
Intent-classified router, the front door to OrchestKit and the DEFAULT entry point for any goal-shaped request. Classifies a plain-English goal and routes it to the right specialist skill. Routing is never overhead, so use it even when the target skill seems obvious; skip only when already executing inside another…
kill-mutants
Analyze surviving mutants from a mutation testing run and write targeted unit tests to kill them. Re-runs mutations to confirm kills.
dream
Nightly memory consolidation — prunes stale entries, merges duplicates, resolves contradictions, rebuilds MEMORY.md index. Use when memory files have accumulated over many sessions and need cleanup. Do NOT use for storing new decisions (use remember) or searching memory (use memory).
mutate
Run mutation testing to verify test quality. Builds a custom mutation tool or uses an existing framework, runs mutations, and reports the mutation score with surviving mutants.