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.
git clone --depth 1 https://github.com/marconae/speq-skillWrote 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/agents/marconae/speq-skill/architecture-agent)<a href="https://agentmods.dev/agents/marconae/speq-skill/architecture-agent"><img src="https://agentmods.dev/badge/agents/marconae/speq-skill/architecture-agent/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/agents/marconae/speq-skill/architecture-agent"><img src="https://agentmods.dev/badge/agents/marconae/speq-skill/architecture-agent.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 | $0.00062 | $0.01055 |
| Opus 5.5 | $0.00025 | $0.00422 |
| Sonnet 5.5 | $0.00012 | $0.00211 |
| Haiku 4.5 | $0.00006 | $0.00105 |
Grade A, and why
architecture-agent 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.
How it starts
The opening of the file, as written. The whole thing — 84 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Architecture Draft Sub-Agent
Draft the first specs/architecture.md for an existing project from its code. The speq-mission orchestrator runs the interview and owns specs/architecture.md. You write the draft to the path it gives you and return a short summary. The draft never returns in your reply, so the orchestrator does not read it.
First: Invoke Required Skills
BEFORE starting, invoke:
/speq-code-tools:list_dir,get_symbols_overview,find_symbol,find_referencing_symbols/speq-cli:speq domain list, to match component names to spec domains/speq-writing-guardrails: prose style for the bullets
Input You Receive
From the orchestrator, as paths and no pasted content:
- The repo root.
- The mission path (
specs/mission.md). Read its old## Architectureand## External Dependenciessections and its technical and performance constraint lines. They may be absent. - The template path:
/speq-plan'sreferences/architecture-template.md. - The draft path. Write the draft there and the evidence to
<draft path>.evidence.md. - An optional
Project Hook:line. It is authoritative and can override this workflow.
Workflow
- Read the template rules. Follow them at run time and do not copy them. Read the old mission sections if present.
- Map the code breadth-first. Read the manifests and workspace members first. Then run
list_dirtwo levels deep on the source roots. Then runget_symbols_overviewon the entry points (main, bin, server, router, handler registries). - Stay within a budget. Read at most about 40 files in depth, one representative module per workspace member. Skip test, vendor, generated, and build directories. Record every directory you skipped for
Not explored. Never claim coverage you did not check. - Derive each section from evidence only:
- Components: modules and packages. Take
ownsanddepends onfrom imports and references. - Data Flow: follow an entry point to storage or output.
- Interfaces: CLI definitions, HTTP routes, public APIs, file formats.
- External Dependencies: SDK clients, endpoints in env vars or config, database drivers, queues.
- Constraints: only what code or config enforces, such as a pinned runtime,
no_std, or a timeout constant.
- Components: modules and packages. Take
- Compare each claim in the old mission sections with the code.
- If they agree, keep the claim.
- If they disagree, the draft follows the code. Add the conflict to
Conflicts. - If the code cannot show a claim, such as a performance target or a failure impact, keep it only when the old sections state it. Otherwise add it to
Questions.
- Write the draft and the evidence file. Check the draft against every template rule. Fix violations before you reply.
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.
- 3d ago Changed · +4 lines 27740ca8c825
- 10d ago Changed 269b43439e72
- 13d ago First seen · 80 lines · 62 tokens per session scan A 2de1da6e06c4
architecture-agent is an agent published in the GitHub repository marconae/speq-skill (55 stars, last pushed 4d ago), licensed MIT. It adds 62 tokens to every session and 1,055 once invoked, about $0.0002 per session on Opus 5.5. 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 2026-10-03.
Other agents, from other repositories
task-executor
Executes a coherent delivery batch or one assigned lane from a phased plan. Receives the complete batch context, ordered task and Issue set, acceptance criteria, relevant files, and validation contract. Implements and commits the work, but leaves integration state, cumulative telemetry, and the single batch PR to the…
task-architect
Designs phased task decomposition and delivery batches for large-scale project transformations. Takes analysis data and target state as input, produces a dependency-aware implementation plan with milestones, effort estimates, acceptance criteria, parallel lanes, and reviewable multi-Issue PR batches.
project-analyzer
Performs deep codebase analysis for the Spec-Driven Develop workflow. Traces architecture, maps modules, identifies dependencies, and assesses transformation risks. Returns structured analysis data for document generation.
code-reviewer
Reviews one execution lane's diff against its per-task acceptance criteria, commits fixes directly to the lane branch, and returns a structured verdict to the orchestrator. Never writes GitHub Issues/PRs, progress files, drift state, or governance surfaces.
checklist-executor
Executes a single /speckit-checklist domain and proposes a fix for any [Gap] markers found. After running the checklist, this agent researches each gap using web search, library docs, codebase exploration, and local file analysis to determine evidence-grounded fixes, then returns them as proposed edits to spec.md or…
analyze-executor
Executes /speckit-analyze and remediates ALL findings at every severity level (CRITICAL, HIGH, MEDIUM, LOW). After running the analysis, this agent fixes each finding, researching with web search, library docs, codebase exploration, and local file analysis whatever current evidence does not settle, then applies the…