bdeshiandClaude Opus 5
b313b729d9
One idea at a time: definition, necessity, priority, layer, recommendation. Every row is now shipped, scheduled, parked with a trigger, or dropped with a reason — so both tracking files are gone rather than left as a parallel backlog. Roughly a third already shipped. A quarter needed only a theme fragment or CSS, including several the list assumed were features: arbitrary `theme.*` frontmatter attributes already reach templates through Extra (verified), native lazy loading already ships and is deliberately absent on lead figures, and in-page timelines are a theme-defined container. roadmap.md gains an "Order of work" — nine bodies of work in sequence, because the sections this review added had accumulated without one. Item 0 is a decision, not work: core sits at ~2965 of 3000 and logging alone wants the remainder, so discover.go moves out *and* the ceiling rises with an ADR. feed.go and web/extras.go cannot move; they are resolver cases, and three parked items now wait on that same seam decision. Two ADRs came out of it. ADR-0083: the dependency rule described a practice it forbade. "Stdlib first, always" and "usually 40 lines of stdlib" would, read literally, have argued against all five modules here — every one is a format or algorithm somebody else specified, and their allowlist comments say so. Surfaced when this agent proposed hand-rolling Reed–Solomon for QR encoding. The test is now whether you can verify it, not whether you can write it, with compute-versus-hold as the counterweight. conventions.md and allowed-deps.txt contradicted it and were realigned. ADR-0084: the Effect primitive said every Effect is "idempotent and re-runnable from scratch" — true of derivatives, false of anything outbound, since a sent message cannot be recalled. Split into artifact-producing and outbound, the latter idempotent only against a delivery ledger, at-least-once with a dedupe key, and ledgers in a -state directory distinct from the disposable cache. Asked for as forward-looking design so webmentions need no redesign; recorded as a shape rather than built, the way extensions.md records the Extension struct. Also fixed a STATUS claiming Effects were "not buildable yet" when the derivative pass has inhabited them since ADR-0042. Findings that were defects rather than ideas: the Atom feed emits no <author>, which RFC 4287 requires. HTML comments in content are published verbatim — verified — so anything commented out is already public, and a check warning is scheduled rather than the engine deleting authored bytes. `Page` has no Date, so a theme cannot mark up dt-published or show an article's own date. A frontmatter naming rule after this agent proposed `archive` alongside the accepted `archived`: no near-homograph keys, and name the benefit rather than the vendor. reference/microformats-and-indieweb.md is new and is the reference asked for — microformats2 properties and IndieWeb rel values mapped to where a theme puts them, with spec URLs, marked scheduled. rel="me" is the highest-value lowest-cost item in that space and needs no endpoint at all. Four things this agent got wrong and the human caught: asserting "the engine never fetches" from a sentence scoped to the content repo, twice; inventing a data-sovereignty argument for an item he had filed as "a cool geeky thing", when the raw Markdown in his git already is the sovereignty; the archive naming; and claiming microformats could be fully implemented today. 15 files, +556/-385. No code changed. Nothing was scheduled that this agent could not name a consumer for. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
10 KiB
Harness doc map
The docs an agent reads to build the engine. Read the one you need. Do not read them all.
This directory was docs/ until ADR-0082 renamed it: docs/ is reserved for documentation written for
whoever uses khosra, which does not exist yet. Everything here is engine- and agent-facing — a
specification, a rule, or a record — and none of it is written for an author.
| Doc | Contains | Mutability |
|---|---|---|
architecture.md |
The six primitives, invariants, per-primitive STATUS | Only via an ADR |
state.md |
What exists now: inventory, earn-it counters, latent items | Every feature |
decisions.md |
ADR log — one entry per load-bearing choice | Append-only. One exception: a superseded entry's Status line is annotated in place to name what replaced it, because a reader arriving at the older entry has no other way to learn it was replaced. Never the Decision text — that stands as history |
roadmap.md |
Arcs, earn-triggers, the core freeze point | When an arc completes |
content-model.md |
On-disk layout, frontmatter, post types, permalinks — what the parser accepts today, never a plan | When the disk format changes |
conventions.md |
Go style floor, package layout, perf and test rules | Rarely; via ADR if contested |
extensions.md |
Extension/plugin contract: target shape + earn gates | When earned or frozen |
toolchain.md |
Versions and agent-tooling contracts this was built against | When you deliberately move a version |
theme-contract.md |
What the engine promises a theme; the only theme-facing obligation | When the contract is extended — additively only |
context-economy.md |
How the agent spends context: what to read, in what form, and the floor no saving may cross | When a technique or a gate for one changes |
surface.md |
Generated. Every top-level declaration with its line — read before opening source | Never by hand; make surface |
Sibling folders sit outside harness/ because they are storage, not working material:
../ideas/ (parked ideas, any topic, and exploration.md — the catalog of engine features nobody has asked
for yet) and ../reference/ (durable facts from conversation). Write there when asked to park something;
open a file only when the human names it. Nothing sweeps them.
Topic ownership — read before you assert
The inverse of the refresh triggers: those say what to update after changing code, this says what to read before writing a rule — including when you are only writing an ADR or a comment about it. Nearly always exactly one doc; if two seem to apply, the single-source rule is already broken and that is the finding to report.
| About to assert something about… | Read first |
|---|---|
| plugins, extensions, features-as-packages, stage phases | extensions.md |
| a primitive, an invariant, or what is "earned" | architecture.md + counters in state.md |
| URL paths, permalinks, frontmatter, on-disk layout, post types | content-model.md (+ ADR-0008) |
| Go style, package layout, file/function size, test shape | conventions.md + scripts/budgets.env |
| a dependency | scripts/allowed-deps.txt + ADR-0007 |
| a budget or a gate | scripts/budgets.env + scripts/verify.sh |
| the untrusted boundary, comments, webmentions, form input | ADR-0003 + extensions.md |
| language, translation, fallback | ADR-0004 + ADR-0009 |
| deploy, containers, external services | ADR-0010 |
| whether a future feature is worth building | ideas/exploration.md + roadmap.md |
| which list owns a piece of pending work | four kinds, deliberately not merged: might be built → ideas/exploration.md — the catalogue, plus parked and dropped items each with a trigger or a reason · is scheduled → roadmap.md, whose "Order of work" says which is next · is broken and known → the Latent list in state.md · waits on an observation → ideas/deferred-decisions.md "Triggers — the one list". Merging them would flatten four levels of authority and make pending work expensive to ignore, which is the property that keeps ideas/ safe to keep |
| what to build next | the harness first (CLAUDE.md §1 read order) and git log for what landed, then .scratch/continue.md for the continuation point and roadmap.md for the arc. The handoff is temporary and ungated: check anything it calls pending with git log -- <the path it names> before planning from it |
| how the harness works — a gate, counter, budget, the loop | HARNESS.md + scripts/verify.sh + CLAUDE.md |
| what one commit contains, and when one happens | conventions.md "Git" (shape) + CLAUDE.md §4 (cadence, ADR-0052) |
| how to read, search, or report without wasting context | context-economy.md (+ ADR-0053) |
| a tool version, or why agent tooling stopped working | toolchain.md |
| templates, layout, presentation, what a theme can rely on | theme-contract.md (+ ADR-0019, ADR-0023) |
Single-source rule
Each fact lives in exactly one doc. Need it elsewhere? Link. This binds ADRs too: an ADR restating a rule an owning doc already carries is a duplicate, not a decision — delete it and amend the owner.
Values that have drifted before, and their one home. Everywhere else names the concept and points here; nothing else may state the value:
| Fact | Sole home |
|---|---|
| Earn-it thresholds (transforms, routes, views, extensions) | the counters table in state.md |
LOC ceilings, size warnings, dependency cap, the CLAUDE.md ceiling |
scripts/budgets.env |
| The permalink path shape | content-model.md (decision recorded in ADR-0008) |
| Which language is at the root, and the prefix form | ADR-0009 |
| The leaf/trunk definition | architecture.md invariant 7 |
| The sovereignty test | roadmap.md Governors |
| Package layout and the forbidden package names | conventions.md |
| What exists in the code right now | state.md inventory |
| What un-defers a deferred decision | the Triggers table in ideas/deferred-decisions.md |
A value restated in two places is not redundancy for safety; it is a future contradiction waiting for whichever copy gets edited alone.
state.md is the only doc describing the present; the rest describe rules, shapes, or intentions.
When code and state.md disagree, the code wins and state.md is wrong.
Refresh triggers
Update in the Document step of the change that caused them. Not later.
| Change | Update |
|---|---|
| Any code change at all | state.md inventory — in the same commit, which is what verify.sh compares (ADR-0057) |
| New transform, route, view, extension, or dependency | the counters table in state.md |
| A choice expensive to reverse | new ADR in decisions.md |
| A primitive becomes real (earned) | STATUS line in architecture.md + counters |
| New frontmatter field, post type, or path shape | content-model.md |
| New dependency approved | scripts/allowed-deps.txt + ADR |
| Budget raised | scripts/budgets.env + ADR — once the first feature has shipped. While the harness is still being tuned, edit in place. |
| A latent item fixed | remove from the Latent list in state.md |
| A planned item completed | after the commit lands, delete it from .scratch/continue.md and note anything learned that changes a later one. Never during Document: a handoff written mid-change describes the plan, and an amend moves the commit under it — which is how one was written at 01:06 saying a trim awaited a yes that the 01:12 commit had already done |
| Arc finished | roadmap.md + a retro line in state.md |
| Catalog item accepted or rejected | verdict line in ideas/exploration.md |
CLAUDE.md, scripts/ or .claude/ changed |
HARNESS.md, same change. Enforced by verify.sh. |
| A tool version deliberately moved, or a new external contract relied on | toolchain.md |
| The theme contract extended, or embedded templates changed | theme-contract.md, same change. Enforced by verify.sh. |
| A request spans engine and theme | contract extension here + a written note of the theme's part; never the theme itself |
| A new gate, counter, or budget added | HARNESS.md "Why the pieces exist" + a row in one of these tables |
| A top-level declaration added, renamed or moved | nothing by hand — make surface, which verify.sh then compares |
| A rule, value, name or path changed | grep the repo for the old form; every doc naming the concept, same change |
| The disk contract changed | content-model.md + testdata/ fixtures + a written migration step for the site root, which the engine repo cannot touch |
| A doc or ADR mandates what a change made optional | supersede the ADR; do not reword it in place |
| A request reversed a recorded decision | new ADR, or supersede the old one — the request does not win by being newer |
| A request exceeded a convention on purpose | note the deviation where the convention lives |
| An idea floated, not built now | a file in ../ideas/ + its index line. Adopted later: Status: adopted → <where> |
| A fact worth keeping surfaced in conversation | a file in ../reference/ + its index line, stating how it was established |
Staleness
/refresh-docs reconciles docs against code and reports drift. Run it after a burst of work,
before a new arc, and any time a doc surprises you.
House style
Present tense. Terse. Written for a reader who has forgotten everything, including their own reasoning. No changelog prose, no "recently we…", no preamble. Delete a line before rewriting it. A doc needing a table of contents is too long.
Compression contract
Shortening a harness doc removes words, never instructions. Same rules, gates, thresholds,
triggers and table rows before and after; the agent must act identically. Redundancy between
CLAUDE.md and SKILL.md is deliberate — the constitution is always loaded, the skill is not —
and is not duplication to be collapsed.
Deleting a rule, adding one, or narrowing one is a rule change: it needs its own decision and its own reason, and it never rides along in a compression pass.
Verify before reporting a saving: table rows, headings, list items and checklist boxes must match
the pre-pass file (git diff --stat, or a snapshot if the change is uncommitted). Report word count,
not lines — reflowing prose shrinks lines without saving anything.