A field report from bootstrapping Services/Vault: ~20 CLI calls produced
three productive writes, and four of those calls were deliberate failures
sent only to discover a vocabulary.
The worst of it was self-inflicted and new. `repoctl skills install` is the
documented way to make the bootstrap skill reachable, and it writes the
three skill files -- which `init` then reported as INIT_PATH_COLLISION and
redirected to `adopt`. Every skill-driven bootstrap of an empty repository
was pushed onto the adoption path. The collision check exists to protect
content the tool did not author, and identical bytes are nothing to
protect: a path already holding exactly what the plan would write is no
longer a collision, at plan time or at write time, while a file whose
content differs still is. `init` reports only the files it actually wrote.
`--lifecycle` advertised no choices, so its vocabulary was reachable only
by sending a wrong value and reading the refusal; it now carries `choices`
like `--type` does.
`--answers-template` prints a commented answers file naming every key, its
allowed values and its default. The answers keys are not the flag names
(`--type` sets `repositoryType`) and nothing printed the mapping. It lists
`summary` and `namespace` too, which are cheap to answer once and otherwise
survive as generated placeholders in the contract.
`--repo .` proposed the name `adopted-repository`: `Path(".").name` is the
empty string, so the ordinary way to name the current repository fell
through to the generic fallback. Resolve before deriving.
Both skills gained what the report had to discover by hand: that `--plan`
already unwraps a saved result envelope, and that sibling repositories are
where a naming convention comes from.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The stewardship skills were written by exactly one code path: `init`, which
only runs on a new repository. Adopting an existing one starts from the
repository-adoption skill, and nothing could put that skill there -- so the
entry point to the procedure depended on the operator already having it
from somewhere else. Services/Monitoring carries three domain skills of its
own and no stewardship skill; `adopt` lists `.claude/skills` among its
collision paths and installs nothing.
repoctl skills install --repo . # travels with the repository
repoctl skills install --global # follows the operator everywhere
repoctl skills status --repo .
Scope is stated, never guessed: the two answers differ in who ends up with
the skills, and neither is the obvious default.
No plan and no approval hash. That discipline protects a repository's own
content; this writes only tool-owned files, in a fixed subdirectory, whose
content is fully determined by the installed version. The house rule it
does keep is that a write never destroys what it did not author: a skill
whose content differs is reported as SKILLS_MODIFIED and left alone until
`--force` says otherwise, and the check covers every skill before the first
byte is written, so a refusal leaves nothing half-updated.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
`tests/fixtures/positive/` holds four valid contracts. Running `discover`
over the machine's Git directory returned six repositories: two real ones
and those four, with nothing in the output to tell them apart. Any
catalog-wide question -- which repositories lack an owner, which are
behind the template version -- was answered over a set that included this
project's own test material.
A repository is the root of a git repository. A directory holding
ops/contract.yaml somewhere inside another repository's working tree is
that repository's content: a fixture, an example, a vendored copy. Skip
it, and say so with CATALOG_NESTED_CONTRACT_SKIPPED rather than silently,
so a genuinely misplaced contract is still visible.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Running a full assess -> plan -> apply -> verify against Infra/LXD and
Infra/Docker in throwaway worktrees, and a bootstrap into a scratch
directory, surfaced the same two mistakes repeated across both commands:
a shape was read where evidence was available, and a missing answer became
a silent default rather than a refusal.
adopt:
- classify secret-shaped paths by content, not by filename. A live
credential blocks; a file tracked against the repository's own ignore
rules and a name-only match are warnings; findings name keys, never
values. In Infra/Docker this turned 14 false blockers into 2 true ones
and surfaced 14 files committed despite `**/.env`.
- carry assess findings into the plan and refuse while an error-severity
finding is unresolved. An owner may accept one through
`acknowledgedFindings` with a reason; the acknowledgement enters the
plan and therefore planHash, so approval covers it.
- derive metadata.name from the git remote instead of the clone directory,
and refuse to plan without a confirmed owner (runbook section 9).
- propose content roots from the whole content vocabulary rather than the
winning type's markers, and recognise N sibling directories sharing a
marker file as one file-set root with a selector, as runbook 7.2 already
specified. Authority defaulted rather than derived now asks.
init: apply the guards adopt already had. Answers are validated against
the vocabularies that already existed in constants (name, type, lifecycle,
access) so init can no longer create a repository that fails its own
validate; owners are required; a directory that already holds the planned
paths is refused at plan time and pointed at adopt; and every path from a
plan is confined before the first byte is written -- init wrote outside
the repository from an edited plan.
validate: mask fenced blocks and inline code spans before scanning for
Markdown links. A PowerShell snippet in Infra/Docker's CLAUDE.md parsed as
a link and kept validate red for a reason unrelated to the repository.
paths: allow "." as a whole content-root path, never as a segment, so the
file-set root runbook 7.2 describes can be declared.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
apply_plan/apply_bootstrap_plan required the raw plan object, but the
natural capture workflow (repoctl ... --output json > plan.json) saves
the full result envelope, so following the documented workflow verbatim
failed the hash check. repoctl now unwraps a stored envelope's `data`
automatically (an idea worth keeping from codex/repoctl-l1's cli.py,
which had the same fallback).
Found while walking docs/repoctl.md's adopt and cross-repo examples
live end to end: two more invocation bugs in the doc itself
(`config add-root` takes a positional path, not `--path`; `--output`
must follow `config` itself, not the add-root/remove-root subaction) --
fixed and clarified in place.
docs/repoctl.md walks a human through what repoctl is, the install-once
/ invoke-from-other-repositories model (folding in the framing already
present on the sibling codex/repoctl-l1 branch's docs/repoctl.md), each
command with real sample output, the exit-code table, and troubleshooting.
docs/specification.md stays the normative machine reference.
kdb/kdb.yaml routes the new document into the tooling-implementation
topic; README.md links it and documents running tests-universal.