Full-text search across the pm items of public GitHub repositories — continuously re-indexed.
Browsing 4,711 pm items from 23 public repositories
The canonical consumer launcher statically imports pm-ops/merge-driver, but pm-ops is a devDependency. npm install --omit=dev with scripts enabled (a source checkout) fails with ERR_MODULE_NOT_FOUND in prepare before any
Pin @unbrained/pm-cli 2026.9.21, pm-changelog 2026.9.18, pm-ops 2026.9.18 exactly; replace the vendored scripts/prepare-merge-driver.mjs with the thin TypeScript launcher over pm-ops/merge-driver; add canonical lint (pm-
Wave 2026-09-22, companion epic pm-cli-website-5s6z. pm-ops IS the canonical source: it already ships the launcher prepare-merge-driver.ts, pm-ops/eslint and pm-ops/duplication, so this wave for pm-ops is (a) pin @unbrai
Certify this package against pm CLI 2026.9.21 and replace the vendored merge-driver plus the missing lint and duplication gates with the canonical pm-ops 2026.9.18 launchers. Companion epic pm-cli-website-5s6z. Open issu
The release workflow treated npm's eventual visibility and the bun mirror as immediate, so a publish the registry accepted a few minutes late was reported as failed and the GitHub Release step was skipped on bun mirror l
Pin @unbrained/pm-cli to 2026.9.21, pm-changelog to 2026.9.18 and pm-ops to 2026.9.18 so this public package certifies the current CLI and consumes the canonical merge-driver, ESLint and jscpd gates instead of a vendored
Add mutation testing for the SDK core so nominal 100 percent coverage is proven to detect behavioral defects, with a ratcheted score and explicit survivor ownership.
Replace mass-generated filler docstrings with concise semantic documentation and add a ratcheted lint or census so coverage remains complete while token cost and misleading boilerplate decline.
Measured 2026-07-27 from the workflow runs API. Across the full recorded history of event=schedule runs of auto-release.yml the conclusions are 52 success and 32 failure, a 38 percent failure rate, with failures on 2026-
Scheduled Auto Release 35702036265 failed coverage because the real tracker integration test requests version 9999.1.1 and relies on an ambient Unreleased section removed during release preparation. Make the fixture own
Deliver the remaining eleven selected release and measurement entrypoints under pm-p5u6: test runner, release preparation, installed-agent acceptance, published verification, compatibility, isolated regression control, c
Adopt fast-check 4.10.2 and tsx 4.23.15 after a disposable admission probe with the repository-pinned pnpm 11.10.0 and unchanged workspace supply-chain settings succeeds. Preserve the Node floor and separately tracked in
Real package-first acceptance passed but a detached telemetry worker recreated its deleted root with global settings and telemetry state. Disable telemetry in this synthetic acceptance environment so its processes cannot
ecosystem review 2026-07-06 hit a full /tmp user quota (tmpfs usrquota, EDQUOT) that broke the local agent shell entirely. Post-mortem inventory: pm-cli-node-compile-cache-1000 had grown to 1.7GB (src/cli.ts enableCompil
Fleet wave 2026-09-22 (companion epic pm-cli-website-5s6z). Pins @unbrained/pm-cli 2026.9.21, pm-ops 2026.9.18 and pm-changelog 2026.9.18 exactly. The prepare hook becomes a thin launcher over pm-ops/merge-driver instead
The three calendar shortcut commands become builtin-calendar package-owned commands (or pm create --template sugar), removing 3 top-level core entries while keeping every flag; core keeps alias shims while the deprecatio
GitHub issue #1284 demonstrates that scripts/generate-release-notes.mjs consumes one budget-truncated list-all JSON page and falsely reports no closed work. Request complete input and reject incomplete or malformed recei
Audit and reconcile stale linked-file rows plus currently unowned source/test artifacts reported by pm validate --check-files. Use pm CLI validation hints and linked-artifact commands so the tracker reflects current code
Inter-file half of the 2026-07-12 too-many/too-long directive (intra-file LOC half = pm-yh8r): regroup high-fan-out directories (src/cli/commands 72, tests/unit/commands 60, tests/integration 32, scripts 23, src/cli 19)
2026-07-04 code-quality sweep: the commands barrel re-exports ~60 runtime symbols + ~80 types that no consumer imports via the barrel (package.json exports never exposes it; only ., ./sdk*, ./cli). knip reports 115 unuse