docs: add agent and domain guidance
Validate and deploy production / validate (push) Canceled after 0s
Validate and deploy production / deploy (push) Canceled after 0s
Validate and deploy staging / validate (push) Successful in 33s
Validate and deploy staging / deploy (push) Canceled after 0s

This commit is contained in:
jb
2026-08-26 21:48:07 -04:00
parent 1fba43272a
commit fdbf811808
5 changed files with 468 additions and 1 deletions
+60
View File
@@ -0,0 +1,60 @@
# Domain docs
Engineering skills must consult this repository's domain documentation before
exploring code.
## Read before exploring
- Read `CONTEXT.md` at the repository root.
- If `CONTEXT-MAP.md` exists, use it to find the `CONTEXT.md` files relevant to
the task.
- Read ADRs under `docs/adr/` that affect the area you will change. In a
multi-context repository, also check `src/<context>/docs/adr/`.
If these files do not exist, proceed without comment. Do not suggest creating
them before the project resolves domain terms or architectural decisions. The
`/domain-modeling` skill, reached through `/grill-with-docs` and
`/improve-codebase-architecture`, creates them when needed.
## File structure
This repository uses the single-context layout:
```text
/
├── CONTEXT.md
├── docs/adr/
│ ├── 0001-event-sourced-orders.md
│ └── 0002-postgres-for-write-model.md
└── src/
```
A repository with multiple bounded contexts can use this layout:
```text
/
├── CONTEXT-MAP.md
├── docs/adr/
└── src/
├── ordering/
│ ├── CONTEXT.md
│ └── docs/adr/
└── billing/
├── CONTEXT.md
└── docs/adr/
```
## Use glossary terms
Use the terms defined in `CONTEXT.md` in issue titles, refactor proposals,
hypotheses, and test names. Do not substitute synonyms that the glossary
rejects.
If a needed concept is absent, reconsider whether the project uses that concept.
If it represents a real gap, record it for `/domain-modeling`.
## Flag ADR conflicts
Call out any proposal that conflicts with an ADR:
> _Contradicts ADR-0007 (event-sourced orders), but worth reopening because…_
+106
View File
@@ -0,0 +1,106 @@
# Issue tracker: Gitea
Issues and specs for this repo live as Gitea issues. Use the
[`tea`](https://gitea.com/gitea/tea) CLI for all operations.
## Repository context
Run commands inside the clone so `tea` can infer the repository and login from
its Git remote. If the clone has multiple Gitea remotes or logins, pass
`--remote <name>` or `--login <name>`. Use `--repo <owner>/<repo>` only outside
the clone.
Before the first operation, verify the instance with
`tea logins list --output json`. If it is missing, ask the user to run
`tea login add`; never ask for or handle their access token.
## Conventions
- **Create an issue**: `tea issues create --title "..." --description "..."`.
For a multi-line body, use shell command substitution around a quoted heredoc:
```bash
tea issues create --title "..." --description "$(cat <<'EOF'
Multi-line issue body.
EOF
)"
```
- **Read an issue with comments**:
`tea issues <number> --comments --fields index,state,author,url,title,body,assignees,labels,comments --output json`.
- **List issues**:
`tea issues list --state open --fields index,state,author,url,title,body,assignees,labels,comments --output json`,
adding `--labels "label-one,label-two"`, `--assignee <user>`, or another
supported filter as needed. Paginate with `--page` and `--limit`; do not
assume the default first 30 results are complete.
- **Comment on an issue**: `tea comments add <number> --description "..."`.
- **Apply or remove labels**: `tea issues edit <number> --add-labels "..."` or
`--remove-labels "..."`. Separate multiple labels with commas.
- **Assign or unassign**: `tea issues edit <number> --add-assignees "..."` or
`--remove-assignees "..."`.
- **Close**: post the closing explanation with
`tea comments add <number> --description "..."`, then run
`tea issues close <number>`.
- **Reopen**: `tea issues reopen <number>`.
- **Raw API fallback**: use `tea api '<endpoint>'` only for a Gitea feature
without a first-class `tea` command. Prefer `{owner}` and `{repo}`
placeholders so repository context remains explicit.
Use `--output json` for machine-readable reads. Do not scrape table output.
## Pull requests as a triage surface
**PRs as a request surface: no.** _(Set this to `yes` if this repo treats
external pull requests as feature requests; `/triage` reads this flag.)_
When set to `yes`, PRs use the same labels and states as issues:
- **Read a PR with comments and diff**:
`tea pulls <number> --comments --fields index,state,author,url,title,body,diff,assignees,labels,comments --output json`.
- **List open PRs**:
`tea pulls list --state open --fields index,state,author,author-id,url,title,body,assignees,labels,comments --output json`.
Determine whether an author is external from the repository's Gitea membership
data; the PR listing does not classify author membership.
- **Comment**: `tea comments add <number> --description "..."`.
- **Apply or remove labels**: `tea pulls edit <number> --add-labels "..."` or
`--remove-labels "..."`.
- **Close**: `tea pulls close <number>`.
Gitea pull requests use the repository's issue index space, so `#42` may refer
to either surface. If the context is unclear, inspect it with
`tea pulls 42 --output json`; if that command reports no PR, read it with
`tea issues 42 --output json`.
## When a skill says "publish to the issue tracker"
Create a Gitea issue with `tea issues create`.
## When a skill says "fetch the relevant ticket"
Run
`tea issues <number> --comments --fields index,state,author,url,title,body,assignees,labels,comments --output json`.
## Wayfinding operations
`/wayfinder` uses one **map** issue with linked **child** issues. These
conventions store relationships in issue bodies, so they work across Gitea
versions.
- **Map**: create one issue labelled `wayfinder:map` containing the Notes,
Decisions-so-far, and Fog sections. Use
`tea issues create --labels "wayfinder:map"` with its title and description.
- **Child ticket**: start the body with `Part of #<map>` and apply a
`wayfinder:<type>` label: `research`, `prototype`, `grilling`, or `task`. Add
each child as a task-list item in the map body to preserve frontier order.
- **Blocking**: put `Blocked by: #<n>, #<n>` near the top of the child body. A
ticket becomes unblocked after every referenced blocker closes.
- **Frontier query**: list all open children, order them by the map's task list,
and discard tickets with an open blocker or an assignee. Select the first
remaining ticket. Fetch each referenced blocker; do not infer its state from
map text.
- **Claim**: resolve the active username with `tea whoami`, then run
`tea issues edit <n> --add-assignees "<username>"` as the session's first
write. Do not assume `@me` works.
- **Resolve**: run `tea comments add <n> --description "<answer>"`, close the
child with `tea issues close <n>`, then append a context pointer and link to
the map's Decisions-so-far section.
+14
View File
@@ -0,0 +1,14 @@
# Triage labels
Engineering skills use five canonical triage roles. This table maps each role to
the label string in this repository's issue tracker.
| Canonical role | Tracker label | Meaning |
| ----------------- | ----------------- | -------------------------------------- |
| `needs-triage` | `needs-triage` | A maintainer must evaluate the issue |
| `needs-info` | `needs-info` | Waiting for more information |
| `ready-for-agent` | `ready-for-agent` | Fully specified and ready for an agent |
| `ready-for-human` | `ready-for-human` | Requires human implementation |
| `wontfix` | `wontfix` | Maintainers will not act on the issue |
When a skill names a canonical role, use the corresponding tracker label.