Skip to content

test(evals): only run eval sources, not their built copies - #1408

Open
JoshuaKGoldberg wants to merge 1 commit into
mainfrom
joshgoldberg/evals-exclude-dist
Open

JoshuaKGoldberg wants to merge 1 commit into
mainfrom
joshgoldberg/evals-exclude-dist

Conversation

@JoshuaKGoldberg

@JoshuaKGoldberg JoshuaKGoldberg commented Oct 5, 2026 •

Copy link
Copy Markdown
Member

The build compiles evals into dist/, which the vitest include glob also matched, so every eval ran twice locally and in CI. This scopes the glob to src/.

The build compiles evals into dist/, which the include glob also matched, so every eval ran twice.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@JoshuaKGoldberg JoshuaKGoldberg changed the title fix(evals): Only run eval sources, not their built copies fix(evals): only run eval sources, not their built copies Oct 5, 2026
@JoshuaKGoldberg JoshuaKGoldberg changed the title fix(evals): only run eval sources, not their built copies test(evals): only run eval sources, not their built copies Oct 5, 2026
@JoshuaKGoldberg
JoshuaKGoldberg marked this pull request as ready for review October 5, 2026 15:08
@github-actions github-actions Bot added the risk: low PR risk score: low label Oct 5, 2026
mr-danya pushed a commit to mr-danya/sentry-mcp that referenced this pull request Oct 6, 2026
…tsentry#1408)

---
Two follow-ups to getsentry#1407 (requested by @BYK on getsentry#1403). Skills stay
opt-out by default per the maintainers' product decision — this does
**not** flip that default.

## 1. Persist the `--no-agent-skills` preference
Passing `--no-agent-skills` once now sticks. `cli setup` records
`defaults.agent-skills=off` in the metadata KV store, and subsequent
runs skip skill installation even without the flag — so a bare `sentry
cli upgrade` (which re-runs `cli setup`) no longer reinstalls skills the
user opted out of.

- new `defaults.agent-skills` metadata key + `getAgentSkillsPreference`
/ `setAgentSkillsPreference` (mirrors the existing `telemetry`
preference)
- `cli setup` persists an explicit `--no-agent-skills` opt-out and
honors a stored opt-out when the flag is absent
- `sentry cli defaults agent-skills on|off` to view/change it; shows up
in `sentry cli defaults`
- re-enable anytime with `sentry cli defaults agent-skills on`

## 2. Skill content quality (issue points 1–3)
Source docs feed the generated skill (`agent-guidance.md`, the `issue`
command fragment, and the `event` field description):
- **points at `sentry issue view <SHORT-ID> --json`** as the fastest way
to get an agent up to speed, in both the workflow example and the field
docs
- **documents the jq shapes** — event data lives under `event.entries[]`
tagged by `type`, not top-level `event.request` (the "first guess that
doesn't match"); added correct recipes for request URL/method and
exception type/value
- **steers away from dumping request/session data** — recommends
`--fields event.id,...` and named-field jq extraction instead of the
whole `event` object, since the `request` entry can carry live
cookies/headers/body

Regenerated `SKILL.md` + `references/issue.md` are included (CI checks
these are in sync).

## Testing
- `vitest run test/commands/cli/defaults.test.ts` — 40 passed (incl. new
agent-skills storage + shape tests)
- `vitest run test/commands/cli/setup.test.ts` — new persistence tests
pass (persists opt-out; honors stored opt-out without the flag). Note: 3
pre-existing agent-skill install tests fail in the sandbox because they
need a live skill fetch/generation that isn't available locally — they
fail identically on clean `main` here and are unrelated to this change.
- `tsc --noEmit` clean
- `biome check` clean on changed files

Refs getsentry#1403
---

---------

Co-authored-by: jared-outpost[bot] <jared-outpost[bot]@users.noreply.github.com>
mr-danya pushed a commit to mr-danya/sentry-mcp that referenced this pull request Oct 6, 2026
…tsentry#1408)

---
Two follow-ups to getsentry#1407 (requested by @BYK on getsentry#1403). Skills stay
opt-out by default per the maintainers' product decision — this does
**not** flip that default.

## 1. Persist the `--no-agent-skills` preference
Passing `--no-agent-skills` once now sticks. `cli setup` records
`defaults.agent-skills=off` in the metadata KV store, and subsequent
runs skip skill installation even without the flag — so a bare `sentry
cli upgrade` (which re-runs `cli setup`) no longer reinstalls skills the
user opted out of.

- new `defaults.agent-skills` metadata key + `getAgentSkillsPreference`
/ `setAgentSkillsPreference` (mirrors the existing `telemetry`
preference)
- `cli setup` persists an explicit `--no-agent-skills` opt-out and
honors a stored opt-out when the flag is absent
- `sentry cli defaults agent-skills on|off` to view/change it; shows up
in `sentry cli defaults`
- re-enable anytime with `sentry cli defaults agent-skills on`

## 2. Skill content quality (issue points 1–3)
Source docs feed the generated skill (`agent-guidance.md`, the `issue`
command fragment, and the `event` field description):
- **points at `sentry issue view <SHORT-ID> --json`** as the fastest way
to get an agent up to speed, in both the workflow example and the field
docs
- **documents the jq shapes** — event data lives under `event.entries[]`
tagged by `type`, not top-level `event.request` (the "first guess that
doesn't match"); added correct recipes for request URL/method and
exception type/value
- **steers away from dumping request/session data** — recommends
`--fields event.id,...` and named-field jq extraction instead of the
whole `event` object, since the `request` entry can carry live
cookies/headers/body

Regenerated `SKILL.md` + `references/issue.md` are included (CI checks
these are in sync).

## Testing
- `vitest run test/commands/cli/defaults.test.ts` — 40 passed (incl. new
agent-skills storage + shape tests)
- `vitest run test/commands/cli/setup.test.ts` — new persistence tests
pass (persists opt-out; honors stored opt-out without the flag). Note: 3
pre-existing agent-skill install tests fail in the sandbox because they
need a live skill fetch/generation that isn't available locally — they
fail identically on clean `main` here and are unrelated to this change.
- `tsc --noEmit` clean
- `biome check` clean on changed files

Refs getsentry#1403
---

---------

Co-authored-by: jared-outpost[bot] <jared-outpost[bot]@users.noreply.github.com>
mr-danya pushed a commit to mr-danya/sentry-mcp that referenced this pull request Oct 6, 2026
…tsentry#1408)

---
Two follow-ups to getsentry#1407 (requested by @BYK on getsentry#1403). Skills stay
opt-out by default per the maintainers' product decision — this does
**not** flip that default.

## 1. Persist the `--no-agent-skills` preference
Passing `--no-agent-skills` once now sticks. `cli setup` records
`defaults.agent-skills=off` in the metadata KV store, and subsequent
runs skip skill installation even without the flag — so a bare `sentry
cli upgrade` (which re-runs `cli setup`) no longer reinstalls skills the
user opted out of.

- new `defaults.agent-skills` metadata key + `getAgentSkillsPreference`
/ `setAgentSkillsPreference` (mirrors the existing `telemetry`
preference)
- `cli setup` persists an explicit `--no-agent-skills` opt-out and
honors a stored opt-out when the flag is absent
- `sentry cli defaults agent-skills on|off` to view/change it; shows up
in `sentry cli defaults`
- re-enable anytime with `sentry cli defaults agent-skills on`

## 2. Skill content quality (issue points 1–3)
Source docs feed the generated skill (`agent-guidance.md`, the `issue`
command fragment, and the `event` field description):
- **points at `sentry issue view <SHORT-ID> --json`** as the fastest way
to get an agent up to speed, in both the workflow example and the field
docs
- **documents the jq shapes** — event data lives under `event.entries[]`
tagged by `type`, not top-level `event.request` (the "first guess that
doesn't match"); added correct recipes for request URL/method and
exception type/value
- **steers away from dumping request/session data** — recommends
`--fields event.id,...` and named-field jq extraction instead of the
whole `event` object, since the `request` entry can carry live
cookies/headers/body

Regenerated `SKILL.md` + `references/issue.md` are included (CI checks
these are in sync).

## Testing
- `vitest run test/commands/cli/defaults.test.ts` — 40 passed (incl. new
agent-skills storage + shape tests)
- `vitest run test/commands/cli/setup.test.ts` — new persistence tests
pass (persists opt-out; honors stored opt-out without the flag). Note: 3
pre-existing agent-skill install tests fail in the sandbox because they
need a live skill fetch/generation that isn't available locally — they
fail identically on clean `main` here and are unrelated to this change.
- `tsc --noEmit` clean
- `biome check` clean on changed files

Refs getsentry#1403
---

---------

Co-authored-by: jared-outpost[bot] <jared-outpost[bot]@users.noreply.github.com>
mr-danya pushed a commit to mr-danya/sentry-mcp that referenced this pull request Oct 6, 2026
…tsentry#1408)

---
Two follow-ups to getsentry#1407 (requested by @BYK on getsentry#1403). Skills stay
opt-out by default per the maintainers' product decision — this does
**not** flip that default.

## 1. Persist the `--no-agent-skills` preference
Passing `--no-agent-skills` once now sticks. `cli setup` records
`defaults.agent-skills=off` in the metadata KV store, and subsequent
runs skip skill installation even without the flag — so a bare `sentry
cli upgrade` (which re-runs `cli setup`) no longer reinstalls skills the
user opted out of.

- new `defaults.agent-skills` metadata key + `getAgentSkillsPreference`
/ `setAgentSkillsPreference` (mirrors the existing `telemetry`
preference)
- `cli setup` persists an explicit `--no-agent-skills` opt-out and
honors a stored opt-out when the flag is absent
- `sentry cli defaults agent-skills on|off` to view/change it; shows up
in `sentry cli defaults`
- re-enable anytime with `sentry cli defaults agent-skills on`

## 2. Skill content quality (issue points 1–3)
Source docs feed the generated skill (`agent-guidance.md`, the `issue`
command fragment, and the `event` field description):
- **points at `sentry issue view <SHORT-ID> --json`** as the fastest way
to get an agent up to speed, in both the workflow example and the field
docs
- **documents the jq shapes** — event data lives under `event.entries[]`
tagged by `type`, not top-level `event.request` (the "first guess that
doesn't match"); added correct recipes for request URL/method and
exception type/value
- **steers away from dumping request/session data** — recommends
`--fields event.id,...` and named-field jq extraction instead of the
whole `event` object, since the `request` entry can carry live
cookies/headers/body

Regenerated `SKILL.md` + `references/issue.md` are included (CI checks
these are in sync).

## Testing
- `vitest run test/commands/cli/defaults.test.ts` — 40 passed (incl. new
agent-skills storage + shape tests)
- `vitest run test/commands/cli/setup.test.ts` — new persistence tests
pass (persists opt-out; honors stored opt-out without the flag). Note: 3
pre-existing agent-skill install tests fail in the sandbox because they
need a live skill fetch/generation that isn't available locally — they
fail identically on clean `main` here and are unrelated to this change.
- `tsc --noEmit` clean
- `biome check` clean on changed files

Refs getsentry#1403
---

---------

Co-authored-by: jared-outpost[bot] <jared-outpost[bot]@users.noreply.github.com>

This branch was successfully deployed

1 active deployment
Actions — eeda48e7 Deployed Oct 5, 2026 by JoshuaKGoldberg via eval #1181
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

risk: low PR risk score: low

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant