AIFlare — Context for your AI-written code, solo or as a team

Overview

Daily Digest is a consolidated summary report for one day's work. Sessions, commits, tag distribution, and key decisions are surfaced on a single page. Run /daily-digest in Claude Code: the MCP tool fetches that day's data, the LLM writes a Markdown report, and the result is saved on the server. The report is private to you in 'My Console' until you publish it to the project console for the rest of the team.

DEV / PM Modes

The same day's data can be summarized through two lenses, chosen by the first command argument (`dev` or `pm`). When omitted, the mode defaults to DEV.

DEV

DEV mode (for engineers)

Keeps technical vocabulary intact — commit hashes, file paths, class and function names — and reports per-session work, tag breakdown, and architectural decisions. Designed for code reviews, 1:1s, and retrospectives.

PM mode (for non-engineers)

Drops technical jargon entirely and re-narrates the day's work in product-facing terms ('checkout screen', 'admin page') framed around business impact. Suited for internal updates, stakeholder syncs, and weekly PM meetings.

Report Structure

Section layout and vocabulary differ by mode. Every section is rendered even when data is sparse — empty sections are replaced with copy like 'limited activity today' rather than omitted, so the structure stays consistent.

DEV

Title — One-sentence summary including the date

Overview — 2–3 sentences capturing total commits, sessions, and focus areas

Session Summary — What each session worked on, the decisions made, and the files touched

Tag Distribution — Breakdown of work by tag — FEATURE / BUGFIX / REFACTOR etc.

Most Changed Files — Files that received the most edits and why

Key Decisions — Notable architectural or design decisions and trade-offs of the day

Title — One-sentence summary including the date (e.g. 'Payment module launched and monitoring beta in progress')

Today's Key Achievements — 2–3 bullets framing progress in user- and business-facing terms

Key Decisions and Their Impact — Re-narrates the day's decisions in non-technical language with a one-line business/product impact

Work Distribution — How effort was distributed by session and area. If work concentrated in one area, that fact is stated explicitly

Reference Statistics — Total commits, sessions, and tag breakdown in a brief appendix tone

PM mode deliberately omits commit hashes, file paths, class/function names, and jargon such as 'refactoring', 'migration', 'endpoint', 'query', 'schema', 'DI', 'JPA'. The LLM is also explicitly blocked from inventing facts that aren't in the data (meetings, customer feedback, schedules, etc.).

Usage Guide

Command Syntax

Type the command below in the Claude Code conversation window. Both arguments are optional — when the mode is omitted, dev is used; when the date is omitted, today's date is used.

/daily-digest [dev|pm] [date?]

Parameters

Name	Type	Required	Default	Description
audience	dev \| pm	Optional	dev	Target reader for the report — dev or pm (case-insensitive). Defaults to dev when omitted.
date	string	Optional	today	Date to query in YYYY-MM-DD format (e.g. 2026-04-09). Defaults to today when omitted.

Argument order: audience → date. Example: /daily-digest pm 2026-04-09. When the mode is omitted, the first argument is interpreted as date.

Execution Flow

1The skill prompt loads, parses the arguments to resolve audience (dev|pm) and date, and runs git log to detect the language used in recent commit messages — the report is written in that language.

2The get_daily_digest tool is called to fetch the day's sessions, commits, tags, and changed files. The response header indicates whether you are in DEV or PM mode.

3Based on the resolved mode, the LLM generates a Markdown report that follows the 6-section DEV spec or the 5-section PM spec. The two are never mixed, and PM mode blocks technical vocabulary and invented facts.

4The report is printed verbatim in the conversation window.

5The save_daily_digest_report tool stores the report on the server together with date, title, content, and audience. If the save fails, the report body has already been printed, so it can be copied directly.

Error Reference

aiflare.yml not found

aiflare.yml is missing from the project root or api_key is empty. See the Getting Started guide to place it.

No data for date

No entries were captured for the specified date. Verify that an AI agent actually committed on that day.

Save failure

The report has already been printed in the conversation. If only the save step failed, copy the printed content for use elsewhere.