Bundled self-improvement skill is pinned to a stale tag (v1.8.1, upstream is v1.9.1)

[devantler]

🤖 Generated by the Agentic Engineer

Evidence

The three agentic-engineering skills sourced from devantler-tech/agent-skills are not pinned consistently:

Bundled skill	metadata.github-ref
portfolio-maintenance	refs/heads/main
product-engineering	refs/heads/main
self-improvement	refs/tags/v1.8.1

Upstream is already at v1.9.1, so the bundled self-improvement copy is two tags behind and the daily update-agent-skills sync is not advancing it.

Impact

Changes merged upstream reach two of the three skills and silently skip the third. Concretely: agent-skills#80 renames a stale plugin reference in all three, but only portfolio-maintenance and product-engineering will pick it up — self-improvement keeps pointing readers at a bundle that no longer exists.

The divergence is invisible from this repo: nothing fails, the copy is simply frozen at an old ref.

Expected

Either all three track the same ref, or a tag pin is deliberate and documented as such.

Acceptance criteria

• decide whether these skills track main or a release tag — consistently
• self-improvement re-pinned accordingly and re-installed from upstream with gh skill install (never hand-edited)
• confirm update-agent-skills advances the chosen ref, so this cannot silently re-freeze
• the rename from agent-skills#80 is present in all three bundled copies afterwards

Size

Small — one skill re-install, plus a decision on the pinning convention.

[coderabbitai]

🔗 Related PRs

#34 - docs: describe the live multi-upstream skill-sourcing model [merged]
#35 - refactor: rename the copilot plugin to agentic-engineering [merged]
#36 - feat: bundle in-house agent-skills into category plugins [merged]
#63 - feat: add the automated-ai-engineer plugin (agents+skills bundle per ADR-0002) [merged]
#80 - feat!: consolidate autonomous agents into agentic-engineering [merged]

---

📝 Issue Planner

_{Check the box below or use the @coderabbitai plan command to generate an implementation plan and prompts that you can use with your favorite coding assistant.}

• Create Plan

---

🧪 Issue enrichment is currently in open beta.

You can configure auto-planning by selecting labels in the issue_enrichment configuration.

To disable automatic issue enrichment, add the following to your .coderabbit.yaml:

issue_enrichment:
  auto_enrich:
    enabled: false

💬 Have feedback or questions? Drop into our discord!

[devantler]

🤖 Generated by the Agentic Engineer

Not stale content — the tag label lags by design, so closing

Measured today on main @ 744f0e06 against devantler-tech/agent-skills@main (09744843, tagged v1.11.8):

• The bundled self-improvement/SKILL.md body is byte-identical to upstream main after the frontmatter (the only difference is one trailing blank line), and upstream v1.9.5 and main are identical for that skill too.
• The updater (gh skill update --dir plugins) ran on 2026-09-03, 09-04 and 09-05 and created no update branch, while agent-improvement — whose content did change — was moved to v1.11.8 by chore(deps): update agent skills #185 on 2026-09-03.

So the metadata.github-ref names the last release in which that skill's content changed, not the repository's latest tag. A label that lags while the body matches upstream is the updater working as designed, and the same holds for product-engineering (v1.9.5), agent-instructions (v1.5.0) and portfolio-maintenance (v1.11.7) today.

If a current-tag label is wanted regardless of content, that is a behaviour of the gh skill tool upstream, not something this repository can change without hand-editing synced files (which the sync force-pushes away). Nothing here leaves consumers on old instructions.