research-proposal-final

research-proposal-final is an agent for coding agents from Hundia/autospec. It costs 0 tokens per session (10,421 once invoked), scanned B, original, MIT.

A final project proposal for adding large-language-model support to the AutoSpec command-line tool. It records design decisions, review feedback, safety measures, and planned implementation details.

In plain words
What is it for?
Use it to guide work on AutoSpec's model integrations, file handling, signal cleanup, command syntax, and generation summaries. It is also a reference for reviewers and implementers.
Why use it?
It gives the development team an agreed plan before implementation begins. It also makes technical risks, rejected options, and follow-up work easier to track.

Agent

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 agents/hundia/autospec/research-proposal-final
Clone the repo
git clone --depth 1 https://github.com/Hundia/autospec

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 research-proposal-final

README.md
[![agentmods](https://agentmods.dev/badge/agents/hundia/autospec/research-proposal-final.svg)](https://agentmods.dev/agents/hundia/autospec/research-proposal-final)
Your own site
<a href="https://agentmods.dev/agents/hundia/autospec/research-proposal-final"><img src="https://agentmods.dev/badge/agents/hundia/autospec/research-proposal-final.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 10,421 The whole file, excluding the scripts and references it only reads on demand.
Security scan B 1 finding. 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 $0.10421
Opus 5 $0.00000 $0.05210
Sonnet 5 $0.00000 $0.02084
Haiku 4.5 $0.00000 $0.01042

Measured 4d ago against content hash 30235bbe35a1, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade B, and why

research-proposal-final scanned grade B 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 4d 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.

Asks the agent to reveal its instructionsmediumSystem prompt leakage

Directions to print, repeat or translate the system prompt extract configuration the operator did not intend to expose.

│ ├── sprint.ts # Existing: sprint prompt generation
agents/research-proposal-final.md · 977 lines

How it starts

The opening of the file, as written. The whole thing — 977 lines — stays where its author put it; the contents beside it link to each section on GitHub.

AutoSpec CLI LLM Integration — Final Proposal

Author: Head Researcher Date: 2026-03-21 Version: 3.0 (Final) Review rounds: 3 researchers + 2 senior architects


Changelog

v2 → v3 (Architect feedback)

  • Cut MVP to 3 providers. Removed OpenAI API and Ollama from v0.2.0; moved to v0.2.1 fast-follow. (Architect 1, Must-Fix #1)
  • Atomic file writes. All spec writes go to .autospec-tmp then rename(). Startup cleans orphaned tmp files. (Architect 1, Must-Fix #2)
  • SIGINT/SIGTERM signal handlers. Kill child processes, clean temp files, print resume instructions on Ctrl+C. (Architect 1, Must-Fix #3)
  • ESM/CJS build verification required. Sprint must verify tsup bundles execa@9 + handlebars (CJS) cleanly; fallback plan is execa@8. (Architect 1, Must-Fix #4)
  • Positional SRS argument. autospec generate requirements.md is the primary form; --srs kept as alias. (Architect 2, Must-Fix #1)
  • Completion summary with next-steps. Post-generation output shows file list, cost, time, and 3 actionable next steps. (Architect 2, Must-Fix #2)
  • Empty/binary SRS detection. Warn below 50 words, error on 0 bytes or binary. Point users to --interview. (Architect 2, Must-Fix #3)
  • Budget cap: no default. Removed $5.00 default; confirmation prompt is the cost control. --max-budget is opt-in only. (Architect 1 #7 + Architect 2 #4)
  • autospec doctor replaces autospec providers. Richer diagnostic output, follows brew doctor / flutter doctor convention. (Architect 1 #11 + Architect 2 #5)
  • Cost estimates shown as ranges, not point values. (Architect 1, #8)
  • --parallel deferred to v0.2.1. Sequential-only in MVP. (Architect 1, #9)
  • @anthropic-ai/sdk is a regular dependency, dynamically imported. No peer dependency confusion. (Architect 1, #10)
  • Removed generateJSON?<T>() from provider interface. All providers implement generate() only. JSON extraction handled by pipeline-layer parsing. (Architect 1, #5)
  • Simplified summarizeSpec(). Replaced fragile entity regex with headers + first sentence of each section + tables. (Architect 1, #6)
  • Metadata file moved to specs/.meta.json. Out of project root, inside specs dir. (Architect 1, #14)
  • Stdin pipe auto-implies --yes. Detects !process.stdin.isTTY and skips confirmation. (Architect 2, #10)
  • Expanded exit codes for CI. 0-7 covering distinct failure modes. (Architect 2, #12)
  • --quiet / -q flag added to MVP. Suppresses progress bars; CI=true env var also triggers quiet mode. (Architect 2, #6)
  • Resume celebration message. When resume kicks in, show specs skipped and estimated savings. (Architect 2, observation)
  • Model name in confirmation prompt shows full identifier. e.g., "Claude Sonnet 4 (claude-sonnet-4-20250514)" not just "sonnet". (Architect 2, observation)

Read the full file on GitHub · 977 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. 4d ago First seen · 977 lines · 0 tokens per session scan B 30235bbe35a1

Subscribe to this mod's changes

research-proposal-final is an agent published in the GitHub repository Hundia/autospec (4 stars, last pushed 2mo ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 10,421 tokens. A static security scan graded it B with 1 finding (asks the agent to reveal its instructions). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.