- Home
- Skills
- Agents & Orchestration
- transcript decisions
transcript decisions
Extract decisions and constraints from transcripts into a structured, agent-readable project record.
Free
Works with the AI Agents you already use
transcript decisions
Example session with this skill installed
I just finished a sync. Transcript is at logs/sync-2024-11-05.txt. We decided to use Postgres instead of Mongo, but we're still not sure about the hosting provider. Update the decisions.
- Read your context and instructions
- Compiled the transcript decisions
I have analyzed logs/sync-2024-11-05.txt and compared it against the existing records in DECISIONS.md. Below is the proposed update for your approval.
Proposal
Keep
- 2024-11-05 | decision | Use PostgreSQL for the primary database. | from: logs/sync-2024-11-05.txt
Supersede
- 2024-09-12 | decision | Use MongoDB for flexible schema. → superseded 2024-11-05 by "Use PostgreSQL for the primary database."
Open
- 2024-11-05 | question | Which hosting provider will manage the production database? | from: logs/sync-2024-11-05.txt
Drop
- "Maybe we look at serverless later." — implementation detail / not agreed
Next steps
- Reply with "Approve" to write these updates to
DECISIONS.mdand ensureAGENTS.mdis updated. - Reply with specific changes if any wording needs adjustment.
- Once approved, I will also check if
DECISIONS-HISTORY.mdneeds a split due to file size.
Connects securely to your tools. The creator never sees your data.
About this skill
Open-source light take on Accord Book: turn a call or chat transcript into an approved DECISIONS.md that Cursor agents actually follow. Full product: https://vector-intelligence.io/accord-book/ Managing technical debt and shifting project requirements is difficult when key agreements are buried in long meeting transcripts. This skill extracts specific decisions, constraints, and unresolved questions from your conversation logs and syncs them directly into a project documentation system designed for AI agents.
What it does
- Extracts agreements by identifying settled decisions and hard constraints from raw text or transcript files.
- Manages state by tracking which decisions are currently in force and which have been superseded by new information.
- Flags uncertainty by logging unresolved questions as open items that require follow-up.
- Agent synchronization by updating a central
AGENTS.mdfile to ensure LLMs reading the repo follow the latest project rules. - History pruning by automatically moving old logs to a history file when the main record grows too large.
How it works
- Source analysis by reading the provided transcript path or chat export.
- Context retrieval of existing decision logs, scanning current sections in full and historical logs from the most recent entries.
- Draft proposal presented in chat showing proposed additions, supersessions, and items to drop.
- Interactive approval where you confirm, edit, or reject specific lines before any files are modified.
Frameworks & tools
This tool manages standard Markdown files (DECISIONS.md, DECISIONS-HISTORY.md, AGENTS.md) within any Git repository. It utilizes wc, grep, and file read offsets to handle large files without blowing out the context window.
Why this beats prompting it yourself
Generic prompts often hallucinate implementation details as "decisions" or fail to properly retire old constraints. This skill follows a strict state management protocol that prevents conflicting instructions from confusing your coding agents.
Use cases
- Syncing standups into a living record of project constraints.
- Updating project scope after a client call or stakeholder meeting.
- Onboarding new agents by giving them a single source of truth for behavioral rules.
- Audit trailing why a specific architectural choice was made months ago.
Known limitations
It requires a clear repository context or path to function. Implementation details and small talk are ignored, as the tool strictly focuses on binding project rules.
Requires a defined repo root. Does not process implementation details. Limits historical scanning to the last 400 entries to manage context.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 10 days ago
- Free with an Agensi account
Requires