Durable documents link into agile content

Table of Contents

This page is a capture in the inbox bucket of the product backlog — a pre-sprint idea, not yet pulled into a sprint as a story.

1. What

Seventy documents under doc/knowledge/, doc/meta, doc/llm/ and doc/recipes/ carry [[id:]] links whose targets are stories, tasks, sprints, versions or captures. Around 140 across some 70 files at the time of writing; compass lint reports the current count, which moves as main merges. The rule is that links run one way: a story links to the durable document it produced, and the durable document does not link back.

The spread is uneven and worth knowing before scoping. doc/knowledge/architecture/history_diff_architecture.org holds nine on its own. Nine recipes under doc/recipes/codegen/ point at one story, Refactor ores.codegen C++ generation. The glossary points at a task, two sprints, a version and a capture. Eleven SKILL.org files point at Clean up the compass work lifecycle.

2. Why

A durable document states current or target state. A link into a sprint's working record carries the reader into a version of the past, and an agent grounding itself follows it and loads material describing a state that no longer holds.

The work is not a sweep of deletions, which is why it needs a person rather than a regular expression. Most of these links carry provenance: the story that introduced a concept, the task that found a bug. Where the link is the only record of why the durable page says what it says, that reason belongs in the page, and the link is removed once it has been absorbed. Where the link adds nothing the page does not already state, it is simply removed.

The lint added alongside this capture reports the violations, so the sweep has a worklist and a regression guard. It is advisory at first, since it fails on 70 files today, and binds once they are clear.

3. References

4. See also

Emacs 29.3 (Org mode 9.6.15)