#Reports

Reports is what happened in a project over a period, measured. The dashboard is a board you compose; Reports is fixed, and every figure says what it counts behind an info mark.

Every project has Reports. Each tab appears with the capability behind it, so a project with everything off still has Summary and Team. Reading a report needs the stats:read permission.

#The tabs

Tab	Needs	Answers
Summary	-	Every capability's period against the one before, with a link to explain it
Flow	Work	How work moved: throughput, cycle and lead time, WIP, what is aging
Backlog	Work	The open work as it stands: by area, tag, priority, size, age, blockers
Planning	Work	Epics and releases: timeline, burn-up, scope change, forecast, slip
Support	Support	Feedback from intake to shipped
Knowledge	Knowledge	Folio activity, stale pages, orphans, the link graph
Delivery	Apps	The four DORA keys, reliability, and the quality CI pushed
Team	-	Load and contribution per member, how work arrives, agents, automations

#The filter bar

The bar sits above every tab and lives in the URL, so a report can be linked and moving between tabs keeps your filters. The default is the last 30 days.

Period - 7, 30 or 90 days, 6 months, a year, all time, or a custom range.
Grain - daily, weekly or monthly, picked from the period's length unless you choose. Weekly starts on Monday.
Compare to previous - the same span just before, as deltas.
Area, epic, release, tag, priority, member - narrow the quests.

A tab shows the filters it does not honour as disabled, rather than quietly ignoring them. Team honours the period alone, because its activity data carries no area or epic. A tag or a segment of a chart links to the quest list narrowed the same way; that list has no period, epic, priority or member filter, so it can show more than the chart counted. Sections with a table or a chart offer CSV.

#What the figures count

Every report reads the same quests: shelved quests are out, and so are the open quests of a draft epic. A completed quest always counts, wherever it was filed.

Figure	Counts	Divides by or notes
Throughput	Quests completed in the period
Cycle time	Median time from accepted to completed	Hold time is taken out
Lead time	Median time from created to completed	Hold time stays: the requester waited
WIP	Quests accepted and not completed, now	Held quests are left out
Net growth	Created minus completed
Waived	Objectives waived	Objectives decided, on completed quests
Progress	Quests done	Total, with shelved quests left out
Scope change	Quests added or cut after the epic started	What it started with
Triage time	Median time from received to accepted or rejected	Reports received in the period
Conversion	Accepted reports whose quests are all completed	Accepted reports
Deploy frequency	Successful deploys per week
Lead time (deploy)	Median time from a quest completed to its first deploy	Matched by release tag or commit
Change failure	Failed deploys and rollbacks	Finished deploys. A rollback is inferred: a deploy back to a tag the copy already ran
Time to restore	Median time from an incident starting to it ending	From monitors
Uptime	Monitor time outside an incident	The period

A median and a p85 go together: half the work took less than the first, and 85% less than the second, so one slow quest moves neither much.

A quest counts under each of its tags, so tag rows overlap and do not add up to the project.

Past states are rebuilt. The cumulative flow reads timestamps and quest history for days before snapshots began, and nightly snapshots after, which are exact.

The forecast is a simulation: a Monte Carlo over the last twelve weeks of the project's throughput, not of the release. Half the simulated futures finish by the date shown.

#Team, and telling agents apart

Team shows what each member carries now (in progress, held, overdue) and what they did in the period: quests created and completed, comments, folio edits, feedback triaged, deploys. It is sorted by name or by load. It never ranks people by output: the columns differ because the work does, and a triager or a folio writer completes no quest and is not idle.

An agent writes as the owner of its API key, so the actor cannot tell the two apart. Every recorded write also carries the channel it arrived on:

Channel	Is
`web`	The browser
`mcp`	An agent over MCP
`cli`	The `lore` CLI, recognised by its user agent, so a best guess
`api`	Anything else with a token
`automation`	An automation rule
`system`	Lore itself
`unknown`	Written before channels were recorded

How work arrives shows the writes per channel, and Agents in the member table is the share of that member's writes that came over MCP.

The rest of the tab:

When work happens - writes per weekday and hour in UTC, over the last 30 days of the period at most, because older activity is kept per day.
Automations - runs by outcome, success rate, duration, and broken rules. Run detail is kept 30 days.
MCP tools - calls per tool that reached this project. A call refused before a permission check names no project, so a stranger's probing shows nowhere.
Explore - the query builder over the project's activity and MCP calls, for the questions nobody anticipated.

Deploys and MCP calls per project are recorded from the release that added them, so earlier periods read zero there.

#The weekly report

The Weekly report automation preset sends Summary's last seven days against the seven before, every Monday at 08:00 UTC, to the member you pick, as an inbox notification or an email. It starts disabled like every preset; see Automations. A preset has one recipient, so reaching several members takes one rule each. A recipient opts out by turning off the automations category in their notification settings.

#For agents and scripts

project_report answers one section for the last N days, the figures as the tab reads them, and lore project report is the same from a terminal:

bash

lore project report summary --days 7

Sections are summary, flow, backlog, planning, support, knowledge, delivery and team. The Write the weekly status agent prompt on the Summary tab builds on it; see Agent prompts.