Guide · Agent skill

How to use the Decision Logger agent skill

Updated 2026-10-29 · written with AI from the project's own documentation and checked against our catalogue data

The decision-logger skill gives an AI agent a memory for decisions made in simulated board meetings. It keeps the full meeting debate separate from the decisions a founder actually approved, so only approved decisions carry forward into future meetings. This guide explains what the skill does, how to install it, how to trigger it with plain requests, and what to check before relying on its output.

Decision Loggeralirezarezvani/claude-skills · MIT Download the skill

Quick start

1Download and unzip. Download the decision-logger ZIP from this page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder.
2Check Python 3. Confirm Python 3 is installed. The skill needs no pip packages.
3Run the demo. Run python scripts/decision_tracker.py --demo to see sample output before using real data.
4Log an approved decision. After a board meeting and founder approval, ask the agent to log the decisions. It writes the raw transcript and checks for conflicts first.
5Resolve any conflicts. If the agent reports a conflict, choose to supersede, merge or defer to the founder.
6Review action items. Run /decision-logger review weekly, and /decision-logger review --overdue to catch missed deadlines.

What is an agent skill, and what does decision-logger do?

An agent skill is a folder containing a SKILL.md instruction file, plus optional scripts and templates, that an AI agent such as Claude can load when a task matches the skill's description. decision-logger is one such skill, from the claude-skills C-level advisor suite.

The skill uses a two-layer memory. Layer 1 stores raw meeting transcripts, including phase contributions, critiques, synthesis and rejected arguments. Layer 2 stores only the decisions the founder approved, along with action items and user corrections. Future board meetings read Layer 2 only, which is meant to stop rejected or unresolved debate from being treated as consensus.

Logs approved decisions in a fixed entry format with owner, deadline, review date, rationale and rejected proposals.
Lists recent decisions and filters them by owner or topic.
Finds action items that are due within 7 days or already overdue.
Flags conflicts: a new decision matching a rejected proposal, two active decisions on the same topic with different conclusions, or the same action assigned to different owners.

Install the decision-logger skill

Option 1 is a local install. Download the decision-logger ZIP from this page and unzip it into ~/.claude/skills/ for all your projects, or into a project's .claude/skills/ folder to limit it to that project.

Option 2 is to upload the ZIP in the Claude apps. Go to Settings, then Capabilities, then Skills, and upload the file there.

The skill's full workflow expects the parent C-Level Advisor and board-meeting skills to be in the same repository. Without them, you can still use the decision-logger commands, but the automatic logging step will not happen.

Local install: ~/.claude/skills/ (all projects) or .claude/skills/ (one project).
Claude apps: Settings → Capabilities → Skills, then upload the ZIP.

Set up the decision folder and requirements

The skill needs Python 3. It uses only the standard library, so there is no pip install step. It stores files in a local folder, ~/.claude/decisions/, which contains raw/ and approved/ subfolders.

The skill's Quick Start runs the bundled script scripts/decision_tracker.py. Run the demo first to see sample output before using it on real decisions.

Requires Python 3 (standard library only).
Storage folder: ~/.claude/decisions/.
Optional for the full workflow: the C-Level Advisor and board-meeting skills in the same repository.

python scripts/decision_tracker.py --demo

Use the commands to review decisions and action items

The skill defines slash-style commands. /decision-logger decisions shows the last 10 approved decisions. Adding --all shows the full history, --owner filters by owner such as CMO, and --topic searches by keyword. /decision-logger review shows action items due within 7 days, and --overdue shows items past their deadline.

The script offers equivalent command-line options for a summary, overdue items, conflicts, owner filtering and search.

/decision-logger decisions: last 10 approved decisions.
/decision-logger decisions --all: full history.
/decision-logger decisions --owner CMO: filter by owner.
/decision-logger decisions --topic pricing: search by keyword.
/decision-logger review: action items due within 7 days.
/decision-logger review --overdue: items past deadline.

python scripts/decision_tracker.py --summary
python scripts/decision_tracker.py --overdue
python scripts/decision_tracker.py --conflicts
python scripts/decision_tracker.py --owner "CTO"
python scripts/decision_tracker.py --search "pricing"

Trigger the skill with natural requests

You do not need to memorise the commands. The skill's description tells the agent when to use it: after logging decisions from a board meeting, when reviewing past decisions, or when checking overdue action items. It is also invoked automatically by the board-meeting skill after the founder approves the synthesis.

Be specific about the meeting, topic or owner. The more the request names the decision, the easier it is for the agent to find the right record.

"Log the decisions from today's board meeting on pricing."
"What did we decide about hiring last month?"
"Show me action items that are overdue."
"Which decisions has the CTO owned this quarter?"

Understand what gets written and how conflicts are handled

When logging, the agent writes the raw transcript to Layer 1, checks for conflicts against the approved index, and surfaces any conflicts for the founder to resolve before writing the approved record. Each approved entry includes a Decision, Owner, Deadline, Review date, Rationale, any User Override, Rejected proposals, Action Items and links to the raw transcript.

Decisions are append-only. Old decisions are superseded rather than deleted, and completed action items are marked done, not removed.

Layer 1 raw files: ~/.claude/decisions/raw/. Archived after 90 days to raw/archive/YYYY/.
Layer 2 approved records: ~/.claude/decisions/approved/, with an index in decisions.md.
Conflicts prompt a choice: supersede the old decision, merge, or defer to the founder.
Rejected proposals tagged DO_NOT_RESURFACE are blocked unless the founder explicitly asks to reopen them.

Safety notes before you rely on it

The skill runs scripts, so read scripts/decision_tracker.py before running it. It does not use the network. Everything is plain Markdown on your local machine, with no audit trail, backup or recovery guarantee. Back up ~/.claude/decisions/ yourself if the records matter.

The skill is not a legal, financial or governance record. Its output should be checked by a human before anyone treats it as authoritative.

Runs scripts: yes. Uses network: no.
Files are local Markdown with no audit or backup guarantees.
Do not use the records as an authoritative legal, financial or governance record.

What people use it for

A founder running AI-assisted leadership meetings keeps a consistent record of what was approved and who owns it.
A small team checks which action items are due this week or already overdue.
A developer studies the two-layer memory pattern to structure their own agent workflows.
A user searches past decisions on a topic such as pricing before a new meeting, so rejected ideas are not raised again without reason.

Frequently asked questions

Is the decision-logger skill free?

The skill is published under the MIT licence, as stated on this page. Check the licence file in the repository for the full terms.

How do I install decision-logger?

Download the ZIP from this page and unzip it into ~/.claude/skills/ for all projects, or into a project's .claude/skills/ folder. In the Claude apps, you can instead upload the ZIP under Settings, then Capabilities, then Skills.

What do I need to run decision-logger?

Python 3 is required, using only the standard library with no pip installs. It stores data in ~/.claude/decisions/. The full workflow also expects the parent C-Level Advisor and board-meeting skills in the same repository.

Does decision-logger use the internet?

No, the source lists network use as no. It runs a local Python script and stores plain Markdown files on your machine.

Can decision-logger replace a note-taking or project management tool?

No. It only handles the board-meeting decision format described in SKILL.md. For general notes or project tracking, use a tool built for that purpose.

Why does the skill keep two separate layers of records?

Layer 1 keeps the full debate, while Layer 2 keeps only founder-approved decisions. Future meetings read only Layer 2, which is meant to prevent rejected or unresolved arguments from being treated as agreed decisions.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

Use decision-logger if you run AI-assisted board or leadership meetings and want a consistent, local record of approved decisions, or if you are studying a two-layer memory pattern for agent workflows. Pick something else if you need a general note or project tool, or an authoritative legal, financial or governance record, since the files have no audit or backup guarantees and outputs need human checking.

Download the skill