docs: drop docs/INDEX.md, add PRD README with format

Remove the one-line docs/INDEX.md (its directory pointers are covered
by docs/README.md's "when to write which document" table). Add
docs/prds/README.md documenting the PRD naming, Status lifecycle, and
section format. Repoint the AGENTS.md repository-layout list at the
new READMEs and add the decisions/ dir.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

AGENTS.md

@@ -28,9 +28,10 @@ the container lifecycle and the copying of skills and env vars into it.
 - `bot-bottle.json` — legacy manifest of named agents (env / skills / prompt
   per agent), consumed by `cli.py`. See "Manifest" under
   "Intended design".
-- `docs/INDEX.md` — pointer to the research notes.
+- `docs/README.md` — docs overview; when to write which document.
-- `docs/prds/` — product requirement docs.
+- `docs/prds/` — product requirement docs (see `docs/prds/README.md` for format).
-- `docs/research/` — research notes (empty for now, kept tracked via `.gitkeep`).
+- `docs/research/` — research notes.
+- `docs/decisions/` — decision records (ADR-lite).
 
 ## Conventions
 

docs/INDEX.md

@@ -1 +0,0 @@
-Research notes live in `research/`. Product requirement docs live in `prds/`.

docs/README.md

@@ -0,0 +1,18 @@
+# Docs
+
+How this project records what it builds and why — and a guide to
+picking the right document for what you're capturing.
+
+## When to write which document
+
+| Artifact | For |
+|---|---|
+| **PRD** (`docs/prds/`) | A feature: what to build, scope, success criteria. |
+| **Research note** (`docs/research/`) | A landscape/tradeoff investigation. |
+| **Decision record** (`docs/decisions/`) | A decision that isn't itself a feature — a policy, a convention, a "we will / won't do this," or a load-bearing choice made inside a larger PRD that deserves to be discoverable on its own. |
+
+A decision that's fully specified by a PRD doesn't need duplicating in
+a decision record. Write one when the *decision* would otherwise be
+buried in prose, lost in an issue thread, or have no in-repo home at
+all (small requests that don't merit a PRD; non-feature choices like
+merge strategy or a trust posture).

docs/decisions/0001-merge-prs-with-rebase.md

@@ -0,0 +1,47 @@
+# ADR 0001: Merge PRs with rebase, not merge commits
+
+- **Status:** Accepted
+- **Date:** 2026-05-28
+- **Deciders:** didericis
+
+## Context
+
+PRs need a merge strategy. Gitea offers merge-commit, squash, rebase,
+and rebase-merge. The project uses [Conventional
+Commits](https://www.conventionalcommits.org/) enforced by a
+`commit-msg` hook, and PRDs typically land as a multi-commit PR where
+each commit is meaningful on its own (e.g. PR #95: a `docs(prd)` commit,
+a `feat(manifest)` implementation commit, and a `docs(manifest)`
+commit). The history should stay readable and the individual
+conventional commits should survive onto `main`.
+
+## Decision
+
+Merge PRs with **rebase** (Gitea's `rebase` style; `Do: "rebase"` via
+the API). The branch's commits are replayed onto `main` with no merge
+commit, producing a linear history that preserves each commit verbatim.
+
+## Consequences
+
+- **Linear history**, no merge bubbles; `git log --oneline` reads as a
+  straight sequence of conventional commits.
+- **Each commit is preserved** (unlike squash, which would collapse the
+  PRD/impl/docs commits into one and lose the staged structure).
+- **Commit SHAs are rewritten at merge.** The replayed commits on `main`
+  get new SHAs, and the source branch is deleted, so a link to a file
+  by *branch name* (`/src/branch/<feature>/…`) dies at merge. This is
+  why links to not-yet-merged files are pinned to a **commit SHA**
+  (`/src/commit/<sha>/…`), which stays reachable via the retained
+  `refs/pull/<n>/head` ref. See
+  `docs/research/issue-tracking-vs-in-repo-decision-history.md`.
+- **Trade-off accepted:** without a merge commit, the "these commits
+  landed together as PR #N" grouping is not recorded in git itself — it
+  lives in forge state (the PR). That is a mild concession against the
+  keep-history-in-the-repo posture; the conventional-commit scopes and
+  PRD references in the messages keep changes traceable without it.
+
+## Links
+
+- `docs/research/issue-tracking-vs-in-repo-decision-history.md` — the
+  commit-pinning consequence above.
+- Observed practice: PRs #92, #93 merged with rebase; #95 to follow.

docs/decisions/0002-agent-identity-claimed-not-vouched.md

@@ -0,0 +1,48 @@
+# ADR 0002: Agent-set git identity is claimed, not vouched
+
+- **Status:** Accepted
+- **Date:** 2026-05-28
+- **Deciders:** didericis
+
+## Context
+
+PRD 0027 lifts `git.user` (name/email) to the agent layer, so an agent
+file may declare its own commit identity. Agent files can live in
+`$CWD/.bot-bottle/agents/` — i.e. they can be supplied by a cloned,
+less-trusted repository. That raises the question of whether a
+repo-supplied agent setting its own git identity is a security concern,
+and whether agent identity should be gated differently for `$CWD`
+agents than for `$HOME` agents.
+
+This record exists because the decision is a **trust posture** worth
+finding on its own, separate from the feature PRD that introduced it.
+The full analysis lives in PRD 0027; the decision is summarized here.
+
+## Decision
+
+Allow agents to set `git.user`, and treat an agent-declared identity as
+**claimed, not vouched**. No `$CWD`-vs-`$HOME` gating on the identity
+field. `git.remotes` stays bottle-only (home-only).
+
+## Consequences
+
+- A cloned repo's agent file can present any commit author name/email,
+  including one that reads like a real person's. This is accepted: git
+  authorship is **not a credential** (push auth is the bottle's remote
+  key/token), is **already forgeable** from inside the bottle at runtime
+  (`git config user.email …`), and was never a trust anchor.
+- If attribution integrity ever matters, the answer is commit
+  **signing** (SSH/GPG), not the author field — so this decision closes
+  no door that was open.
+- `git.remotes` is deliberately *not* lifted to the agent layer: it
+  carries credentials and host trust (IdentityFile, KnownHostKey) and
+  remains a bottle-only, home-only concern.
+- Revisit if a future change ever makes commit identity load-bearing
+  (e.g. enforced signing keyed on author), at which point gating
+  `$CWD`-supplied identities would matter.
+
+## Links
+
+- PRD 0027 (`docs/prds/0027-agent-git-user-identity.md`) — full trust
+  analysis and schema.
+- Issue #94, PR #95 — the feature this decision was made for.

docs/decisions/README.md

@@ -0,0 +1,42 @@
+# Decision records
+
+Short, durable records of decisions — one file per decision. This is a
+lightweight [Architecture Decision Record](https://adr.github.io/)
+practice: capture *what was decided and why* in a versioned file so the
+reasoning lives in the clone, not in a forge issue thread or a chat log
+that disappears when the provider does.
+
+See `docs/research/issue-tracking-vs-in-repo-decision-history.md` for
+the rationale behind keeping decision history in-repo, and
+[`docs/README.md`](../README.md) for when to write a decision record
+vs. a PRD or research note.
+
+## Format
+
+One Markdown file per decision, numbered sequentially and zero-padded
+(`0001-…`, `0002-…`), matching the PRD numbering style. Keep it short —
+the discipline is writing it down, not the ceremony.
+
+```markdown
+# ADR 0000: <short imperative title>
+
+- **Status:** Proposed | Accepted | Superseded by ADR NNNN
+- **Date:** YYYY-MM-DD
+- **Deciders:** <who>
+
+## Context
+What forced the decision; the constraints in play.
+
+## Decision
+What we decided, stated plainly.
+
+## Consequences
+What follows — the good, and the costs/trade-offs accepted.
+
+## Links
+PRDs, research notes, issues/PRs. Forge links are convenience
+pointers; the reasoning above must stand without them.
+```
+
+The records are the index: `ls docs/decisions/` or skim the titles.
+No hand-maintained list to keep in sync.

docs/prds/0025-bottle-extends.md

@@ -39,6 +39,41 @@ trust boundary*: only `$HOME` bottles can declare it, only `$HOME`
 bottles can be its target. Cloned repos still cannot author
 bottle-equivalent config.
 
+## Alternatives considered
+
+The question raised in issue #88 was *where composition should live*.
+Three points in that design space, recorded here so the decision
+stands on its own without the issue thread:
+
+1. **Duplicate bottles (status quo).** Copy `dev.md` to `staging.md`
+   and edit. Zero new mechanism, but every shared field drifts: a
+   route added to `dev` is silently missing from `staging`. This is
+   the pain that prompted #88.
+
+2. **Agent-side `bottle_config:` override (the original #88
+   proposal).** Let an agent file carry an inline block that merges
+   over its referenced bottle. Ergonomically attractive — one file,
+   no second bottle — but it **breaks the trust boundary**: agent
+   files can come from `$CWD/.bot-bottle/agents/` in a cloned repo, so
+   a clone could redeclare egress routes, env mappings, and git
+   remotes — i.e. grant itself bottle-equivalent authority over
+   credentials and network egress. The home-only-bottle invariant
+   exists precisely to stop this.
+
+3. **Bottle-side `extends:` (chosen).** Move composition to the
+   bottle layer, where it inherits the home-only property for free:
+   only `$HOME` bottles can declare `extends:`, and only `$HOME`
+   bottles can be its target. Identical duplication relief to option
+   2, none of its trust erosion. The cost is that an override requires
+   a (home-owned) child bottle rather than an inline agent block —
+   which is the *point*: the override authority stays in `$HOME`.
+
+`extends:` wins because it solves the duplication pain entirely on the
+trusted side of the agent-vs-bottle boundary. (PRD 0027 later lifts a
+deliberately narrow, non-credential field — `git.user` — to the agent
+layer, on the separate reasoning that commit identity is not a
+capability; egress, credentials, and remotes stay bottle-only.)
+
 ## Goals / Success Criteria
 
 - Add `extends: <bottle-name>` to the bottle frontmatter schema.
@@ -58,9 +93,9 @@ bottle-equivalent config.
 
 ## Non-goals
 
-- **No agent-side `bottle_config:`.** That's the design issue #88
+- **No agent-side `bottle_config:`.** Option 2 under "Alternatives
-  considered and weighed against; this PRD is the alternative
+  considered" — weighed and rejected on trust grounds. Don't
-  picked in the issue's design discussion. Don't reintroduce it.
+  reintroduce it.
 - **No additive list merges** (e.g., `routes: append` keyword).
   The `extends:` design uses full-replace for list-valued fields
   (see "Merge rules"); if a use case shows up that genuinely
@@ -167,7 +202,7 @@ Bottles continue to be loaded from `$HOME/.bot-bottle/bottles/`
 only (`Manifest.from_md_dirs` is unchanged). The `extends:` field
 references another file in that same directory. No cwd-readable
 file gains the ability to declare or modify bottle config — the
-attack surface from issue #88's comment thread stays closed.
+attack surface from option 2 ("Alternatives considered") stays closed.
 
 If a future change ever introduces cwd-loaded bottles, the
 `extends:` resolver should be gated to forbid a `$CWD` bottle

docs/prds/README.md

@@ -0,0 +1,63 @@
+# Product requirement docs
+
+One PRD per feature: what to build, why, and how it's scoped. The PRD
+is the durable spec — it should stand on its own without a forge issue
+thread (see [`../README.md`](../README.md) for when a PRD is the right
+document vs. a research note or a decision record).
+
+## Naming and numbering
+
+`NNNN-kebab-title.md`, zero-padded and sequential (`0024-…`, `0025-…`).
+Numbers are never reused; gaps are fine (there is no 0005). The number
+is assigned at creation and stays fixed for the life of the doc.
+
+## Status
+
+The `Status:` line near the top tracks the PRD's lifecycle:
+
+- **Draft** — proposed, not yet shipped.
+- **Active** — the design has shipped to `main` and is in effect.
+- **Superseded by [PRD NNNN](…)** — replaced by a later PRD; kept for history.
+- **Retargeted by [PRD NNNN](…)** — folded into a later PRD's scope.
+
+## Format
+
+```markdown
+# PRD NNNN: <short title>
+
+- **Status:** Draft
+- **Author:** <who>
+- **Created:** YYYY-MM-DD
+- **Issue:** #<n>            # optional — convenience pointer only
+
+## Summary
+One paragraph: what this builds and the pain it solves.
+
+## Problem
+The current state and why it's inadequate.
+
+## Goals / Success Criteria
+Bullets a reviewer can check the finished work against.
+
+## Non-goals
+What this explicitly does not do — and won't, to head off scope creep.
+
+## Scope
+In scope / out of scope, when the boundary needs spelling out.
+
+## Design
+How it works: schema, data flow, diagrams, algorithms as needed.
+
+## Implementation chunks
+Ordered, mergeable steps (optional; for multi-PR features).
+
+## Open questions
+Unresolved decisions — resolve or fold into Design before shipping.
+```
+
+Sections are a guide, not a straitjacket: drop the ones a given PRD
+doesn't need (a small change rarely needs Scope or Implementation
+chunks) and add others where they help (e.g. Testing strategy,
+Alternatives considered, References). Keep the rationale self-contained
+— inline the reasoning rather than linking out to an issue thread, so
+the PRD survives a forge migration.

docs/research/issue-tracking-vs-in-repo-decision-history.md

@@ -0,0 +1,196 @@
+# Tracking feature requests in Gitea vs. in-repo decision history
+
+Research into whether bot-bottle should track feature requests (and the
+decision-making around them) as Gitea issues, given that the project
+already records specs in-repo as PRDs (`docs/prds/`) and rationale as
+research notes (`docs/research/`). The stated constraint is that the
+*history of why we decided things* should be durable and portable —
+not locked into a single forge (Gitea today, conceivably GitHub or
+something else tomorrow).
+
+## Summary
+
+Keep using issues, but demote them. The repository — not the forge — is
+the system of record for any decision you would be unhappy to lose.
+Issues are an excellent **inbox and coordination surface** (cheap
+capture, triage, async discussion, notifications, auto-linking) and a
+**poor archive** (provider-locked storage, brittle numeric references,
+rationale stranded in comment threads). The failure mode to avoid is the
+one already present in the repo: a PRD whose reasoning is only complete
+if you also read a Gitea issue thread.
+
+The fix is a discipline, not a tool: **every load-bearing decision gets
+reified into a versioned file in the repo before the issue that prompted
+it is closed.** PRDs already do this for features; the gap is (a) small
+requests that never merit a PRD and (b) decisions that aren't features
+at all (e.g. "we merge with rebase," "author identity is claimed-not-
+vouched"). Close that gap with a lightweight in-repo decision log. Then
+issues can be as disposable as the forge makes them, and migrating
+forges costs you triage state, not history.
+
+## Why this even comes up here
+
+The project already leans on the repo for durable artifacts:
+
+- **PRDs** (`docs/prds/0001…0027`) — the spec and its rationale.
+- **Research notes** (`docs/research/`) — the "why," with tradeoffs.
+- **Conventional-commit history** — a machine-greppable change log.
+
+But the issue layer has quietly become load-bearing in places:
+
+- PRD 0025 says it picked "option 3" *"from the #88 design
+  discussion"* and that the rejected alternative lives "in issue #88's
+  comment thread." The PRD's rationale is therefore **incomplete without
+  the issue**. If Gitea is gone, the strongest argument for the chosen
+  design is gone with it.
+- PR #89's description links `…/didericis/claude-bottle/issues/88` —
+  the **pre-rename** repo path (the project was Codex-bottle/claude-
+  bottle before the bot-bottle rebrand). That link is already
+  half-dead: a concrete demonstration that forge URLs rot under the
+  most routine event imaginable, a rename.
+- Issue/PR numbers (`#88`, `#90`, `#94`, `#95`) are **forge-assigned
+  from a shared sequence**. They cannot be reconstructed from a clone,
+  and they collide/renumber on import into a different tracker.
+
+So the question isn't academic. The current practice is already
+producing references that don't survive a rename, let alone a migration.
+
+## What each medium is actually good at
+
+| Concern | Gitea issue | In-repo file (PRD / note / log) |
+|---|---|---|
+| Capture friction | Near-zero — file a one-line idea | High — a PRD is a heavy artifact; a note less so |
+| Triage (labels, milestones, open/closed, assignee) | Native, good | Absent / hand-rolled |
+| Async discussion + notifications | Native (threads, @mentions, watch) | None — needs a PR review or out-of-band chat |
+| Auto-linking (`Closes #N`, PR↔issue, commit↔issue) | Native | Manual cross-reference |
+| Version control of the content | None — lives in Gitea's DB | Full — diff, blame, branch, revert |
+| Travels with `git clone` | No | Yes |
+| Survives forge migration | Degrades (export/import; threads, authors, timestamps, refs lossy) | Unaffected |
+| Survives forge outage | Inaccessible | Local clone has it |
+| Greppable offline / by tooling | Only via API | `grep docs/` |
+| Reproducible identifiers | Forge-assigned numbers | Filenames you control (`0027-…`) |
+
+The split is clean: **issues win on the live, social, coordination axes;
+the repo wins on every durability and portability axis.** Nothing about
+that table says "pick one." It says "use each for what it's good at, and
+don't let the durable thing depend on the ephemeral one."
+
+## Lock-in failure modes (the cons, concretely)
+
+1. **Stranded rationale.** The single most valuable output of a feature
+   discussion — *why we rejected the obvious alternative* — usually
+   emerges in a thread and dies there unless someone copies it into the
+   spec. PRD 0025 is already in this state.
+2. **Reference rot.** `Closes #88` / "see issue #90" are meaningful only
+   against one forge instance at one point in time. A rename already
+   broke one such link; a migration would break all of them and
+   silently renumber the survivors.
+3. **Two sources of truth.** A PRD carries `Status: Draft`; the issue
+   carries open/closed. They drift. Which is authoritative?
+4. **Availability coupling.** Self-hosted Gitea down (or the Tailscale
+   path to it down) means the backlog and its history are unreachable,
+   even though the code and PRDs are right there in the clone.
+5. **Export is lossy.** Gitea→GitHub (or the reverse) moves issue *text*
+   tolerably but mangles cross-references, comment authorship for
+   non-mapped users, timestamps, and reactions. The graph of "#88 → PR
+   #89 → commit abc" does not survive intact.
+
+None of these are arguments against *having* issues. They're arguments
+against issues being the **only** place a decision is recorded.
+
+## Pros of keeping issues anyway
+
+Worth stating plainly, because "just use the repo for everything"
+overcorrects:
+
+- A PR per half-formed idea is absurd; issues are the right weight for
+  "someone should look at X someday."
+- Triage state (priority, milestone, assignee, open/closed) is genuine
+  project-management value the repo does not natively provide.
+- Notifications and threaded discussion are how a decision *gets made*
+  before it's ready to be written down. Killing issues doesn't move that
+  conversation into the repo — it moves it into chat/DMs, which is
+  *worse* for durability, not better.
+- `Closes #N` automation and PR↔issue linkage are real ergonomics.
+
+The goal is not to abandon the tracker. It's to make sure that when the
+tracker eventually goes away, you lose the *backlog*, not the *history*.
+
+## What belongs where
+
+- **Gitea issue** — intake, triage, status, and the live discussion.
+  Treat it as a **cache**: useful now, expendable later.
+- **PRD (`docs/prds/`)** — the durable spec for anything that warrants
+  one. Rule: a PRD must be **self-contained**. Synthesize the issue
+  discussion into the Problem / Design / Open-questions sections;
+  reference the issue as a convenience pointer, never as the only home
+  of a load-bearing argument. (Retrofit PRD 0025: inline the #88
+  "option 3 vs `bottle_config:`" reasoning so the PRD stands alone.)
+- **Research note (`docs/research/`)** — the durable "why," exactly like
+  this file. Comparative analysis, landscape surveys, tradeoffs.
+- **Commit message** — the durable "what changed and why, at this point
+  in the diff."
+- **Decision log (proposed, see below)** — durable record of decisions
+  that aren't features and don't merit a PRD.
+
+## Closing the gap: a portable decision record
+
+Two classes of decision currently have no in-repo home:
+
+- **Sub-PRD feature requests** — too small for a PRD, but you still want
+  a tracked "we will / won't do this, because." Today these live only as
+  issues.
+- **Non-feature decisions** — "merge with rebase, not merge-commit,"
+  "agent identity is claimed-not-vouched," "bottles are home-only."
+  Some land inside a PRD that happens to touch them; many are folded
+  into chat and lost.
+
+Options, cheapest first:
+
+1. **An ADR-lite log under `docs/decisions/`.** One short Markdown file
+   per decision: context, decision, consequences, date, links. This is
+   the industry-standard Architecture Decision Record pattern, and it's
+   a near-exact fit for "track decision history, portably." Numbered
+   like PRDs (`0001-merge-with-rebase.md`). ~10 lines each; the
+   discipline is writing them, not the format.
+2. **Reuse the journal.** The repo ships an `init-entry` skill that
+   writes timestamped prose to `docs/JOURNAL.md` (not yet created here).
+   A stream-of-thought journal is a fine home for decision *narrative*
+   and is already part of the toolchain — lower ceremony than ADRs, less
+   structured for later retrieval. The `tag-entries` skill could tag
+   decision entries for grep-ability.
+3. **Periodic issue export.** Belt-and-suspenders: a scheduled job hits
+   the Gitea API and dumps open/closed issues + comments to JSON under
+   `docs/issues-archive/`, committed. Preserves the raw thread against
+   forge loss without changing daily workflow. Mechanical, not a
+   substitute for reifying rationale (a JSON dump of a thread is
+   evidence, not a decision).
+
+These compose: ADRs/journal for the *decision*, optional export for the
+*raw evidence*, issues for *live coordination*.
+
+## Recommendation
+
+1. **Keep Gitea issues for intake, triage, and discussion.** Don't fight
+   the forge on the things it's good at.
+2. **Make the repo the system of record.** Adopt the rule: no decision
+   is "done" until its rationale exists in a versioned file (PRD,
+   research note, or decision log). The issue is a pointer, never the
+   sole source.
+3. **Add `docs/decisions/` (ADR-lite).** Smallest change that closes the
+   real gap — sub-PRD requests and non-feature decisions. Start by
+   back-filling the few decisions already made only in threads or chat
+   (rebase-merge policy; the agent-identity trust call from PRD 0027).
+4. **Retrofit PRD 0025** to inline its #88 rationale, removing the one
+   existing hard dependency on a forge thread.
+5. **Treat issue numbers as disposable.** When a PRD/commit cites an
+   issue, ensure the cited content is mirrored in-repo so the citation
+   degrades to a dead-but-harmless link, not lost information. (The
+   already-broken `claude-bottle/issues/88` link is the warning.)
+6. **Optional:** automate a Gitea issue export into the repo if you want
+   the raw threads preserved without manual transcription.
+
+Net: issues stay, because the alternative to issues is chat, which is
+worse. But the project's durable memory must live where the project
+already lives — in the clone — so that switching forges, or losing one,
+costs you a backlog you can rebuild, never a history you can't.