[release-v1.23.x] Update component versions

[github-actions]

Component version updates

• pac: v0.48.0 -> v0.48.1

---

This PR was automatically created by the auto-update workflow.

[fullsend-ai-retro]

🤖 Finished Retro · ✅ Success · Started 3:28 AM UTC · Completed 3:42 AM UTC
Commit: bbb7b06 · View workflow run →

[fullsend-ai-retro]

Retro Analysis: PR #498 — Bot version bump with no agent activity

PR #498 is a github-actions[bot]-generated version bump (pipelines-as-code v0.48.0 → v0.48.1) on the release-v1.23.x branch. No fullsend agents (triage, code, review, fix) participated in the PR lifecycle. The PR was approved and merged by theakshaypant on 2026-07-21. The only fullsend activity was the retro agent being dispatched on the close/merge event — analyzing a 3-line version bump where no agents ran and there is nothing meaningful to learn from.

This retro found no novel improvements to propose. All findings are covered by existing open issues:

• Evidence for fullsend#5295 and fullsend#5185: PR [release-v1.23.x] Update component versions #498 is a github-actions[bot]-authored PR with zero agent involvement, yet retro was dispatched unconditionally on the pull_request_target/closed event. This is the exact pattern those issues propose skipping. The current hardcoded bot-skip list does not include github-actions[bot].

• Evidence for fullsend#5296: Pattern analysis of the last 15 PRs in this repo revealed that fork PRs fix: remove automated opc version updates #488 and chore: add agents repo to allowed_remote_resources #490 had the Route job fail on close/merge events due to actions/checkout refusing to check out code from fork PRs in pull_request_target workflows. This prevented retro from running on PRs that had actual review agent activity — the opposite of the desired behavior.

• Evidence for fullsend#3551: On PR Update opc version in makefile #495, Route succeeded but retro was skipped because the dispatch role check maps the retro stage to the fullsend meta-role, which was not in the repo's configured roles at the time.

The net effect in this repo is an inverted targeting pattern: retro runs exclusively on low-value bot PRs (where no agents participated) and never runs on high-value fork PRs (where review agents actually provided feedback). This is the combined effect of the three existing issues above.

Additional observations:

• The repo was onboarded to fullsend 18 days ago (PR chore: initialize fullsend per-repo installation #480, 2026-07-03) and is on reusable-dispatch v0.27.0. A pending dependabot PR (Bump fullsend-ai/fullsend/.github/workflows/reusable-dispatch.yml from 0.27.0 to 0.31.0 #500) would upgrade to v0.31.0, which may resolve some routing issues.
• The repo has no CLAUDE.md file, relevant to fullsend#2441.
• Review coverage is low (2 out of 15 recent PRs received reviews), partly due to fork PR authorization and partly by design for bot PRs.