* perf(rls): wrap session GUC reads in RLS policies for per-statement InitPlan
RLS policies read current_setting('session.user' / 'session.groups' /
'session.pgroups' / 'session.folders_read' / 'session.folders_write')
directly inside their USING / WITH CHECK predicates. Postgres treats those
unwrapped calls as potentially row-varying and re-evaluates them once per
scanned row, on the read path of every workspace-scoped table.
The GUCs are set with SET LOCAL (set_config(..., true)) in
set_session_context(), so they are constant for the duration of a statement.
Wrapping each session-derived subexpression in a scalar sub-select lets the
planner hoist it to a one-time InitPlan (evaluated once per statement, reused
for every row) — same rows in, same rows out, N per-row GUC lookups collapse
to 1. Array-producing subexpressions keep an explicit ::text[] cast on the
sub-select so `= ANY (...)` / `?|` stay in their array-operand form rather
than being reparsed as a row-returning subquery.
The consolidating migration recreates every existing policy (across ~30 prior
migrations) whose predicate reads a session GUC, by deparsing the current
predicate and substituting the wrapped forms; the down migration is the exact
inverse (byte-identical round-trip). The adding-a-trigger skill documents the
wrapped form so new trigger tables inherit it.
Surfaced by pgrls (PERF001).
Fixes GIT-919
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* docs(adding-a-trigger): fix RLS example cast placement for = any context
The `= any(...)` example put the ::text[] cast inside the sub-select, which
Postgres parses as a row-returning subquery and rejects at CREATE POLICY with
`operator does not exist: text = text[]`. Move the cast outside the sub-select
(matching the migration's canonical form) so the operand stays in array form,
and note why.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* ci: make local-review a single-source-of-truth cross-agent skill
The repo already had parallel skills directories (.agents/skills/ and
.claude/skills/) drifting between agents. Consolidate local-review
onto one canonical file in .agents/ and symlink the .claude/ entry to
it so Claude Code and Pi share the exact same SKILL.md (Anthropic's
Skills format is supported by both, only the discovery directory
differs).
The canonical SKILL.md now points reviewers at
.github/review-prompt-shared.md as the policy source — same shared
prompt the GitHub auto-review workflows already use — so local
reviews and CI reviews stay in lockstep.
Codex CLI doesn't support repo-level slash commands (its prompts live
in ~/.codex/prompts/). For Codex parity, ship scripts/local-review.sh
which pipes the SKILL + shared policy into 'codex exec' (or 'pi -p'
as a uniform entry point). Update AGENTS.md to document the three
invocation paths.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* ci: make all skills cross-agent — single source in .agents/, symlink .claude/
Turn every skill into a single canonical file under .agents/skills/ and
a symlink under .claude/skills/. Editing any one SKILL.md now updates
all three CLIs (Claude Code reads .claude/, Codex and Pi auto-discover
.agents/).
Per-skill resolution:
- local-review: already symlinked (prior PR #9037)
- rust-backend, svelte-frontend: identical content → symlink, no edit
- refine: only differed in user_invocable frontmatter → add to canonical
- native-trigger: .claude/ had a newer Step 17 (sidebar visibility)
missing from .agents/ → use Claude content as canonical
- commit: .claude/ embedded a Claude-specific Co-Authored-By trailer
the harness already injects automatically → drop from canonical, use
agent-neutral .agents/ version
- pr: generalize "Run /local-review" to "Invoke the local-review skill
(/local-review in Claude Code, $local-review in Codex, pi --skill
local-review in Pi)" and drop the Claude-specific "Generated with
Claude Code" attribution from the PR body template — the harness
that invoked the skill can add its own trailer if desired
- adding-a-trigger: was only in .claude/ → move to .agents/ canonical
- update-sqlx: was only in .agents/ → add .claude/ symlink
Also drop scripts/local-review.sh — wrapper is redundant now that all
three CLIs natively discover the skill from their respective
directories. Update AGENTS.md accordingly.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Ruben FiszelandClaude Opus 4.8