Skip to content

feat: improve skill scores for social-autoposter - #1

Open
rohan-tessl wants to merge 1 commit into
m13v:mainfrom
rohan-tessl:improve/skill-review-optimization
Open

feat: improve skill scores for social-autoposter#1
rohan-tessl wants to merge 1 commit into
m13v:mainfrom
rohan-tessl:improve/skill-review-optimization

Conversation

@rohan-tessl

@rohan-tessl rohan-tessl commented Apr 13, 2026

Copy link
Copy Markdown

Hey @m13v 👋

I ran your skills through tessl skill review at work and found some targeted improvements.

image

Here's the full before/after:

Skill Before After Change
python-cdp-scripts 10% 90% +80%
client-website 80% 93% +13%
posthog-funnel 81% 85% +4%
browser-script 89% 93% +4%

This PR covers 4 of your 8 skills to keep the contribution focused and reviewable.

Changes summary

python-cdp-scripts (10% → 90%)

  • Added YAML frontmatter — the file had no --- block, causing validation to fail entirely. Added name, description with specific actions (navigates pages, extracts data, scrapes stats, audits posts, calls internal APIs) and explicit trigger terms.

client-website (80% → 93%)

  • Split 1307-line monolith into SKILL.md + references/ — extracted component blueprints (Header, Footer, FAQ, case studies, precall page, about page, SVG icons) into references/COMPONENTS.md and typography/color tables into references/DESIGN-SYSTEM.md. Main SKILL.md now serves as a concise orchestration overview (273 lines) with links to detailed references.
  • Improved progressive disclosure score from 1/3 to 3/3.

posthog-funnel (81% → 85%)

  • Improved description with action verbs — changed from passive "Full product funnel stats" to active "Queries PostHog, Neon Postgres, and Cal.com to compile full product funnel metrics." Added 'conversion rates' trigger term.
  • Added error handling section for common failure modes (missing env vars, PostHog API errors, DB connection failures).

browser-script (89% → 93%)

  • Improved description specificity — added concrete actions (navigates pages, clicks elements, fills forms, extracts data) and the key performance benefit (replaces 5-10 calls with 2, cutting tokens 10-20x).

Want your remaining skills optimised too? 🚀

This PR covers 4 of your 8 skills to keep the contribution focused and reviewable. We have tooling that can go further:

  • Optimize remaining 4 skills automatically (same AI-powered pass as above)
  • Add a GitHub Actiontesslio/skill-review-and-optimize — that automatically reviews (and optionally optimizes) any SKILL.md changed in future PRs. Review mode works with zero secrets; maintainers can add TESSL_API_TOKEN for AI suggestions and the /apply-optimize comment flow.

Interested? Just tick the box below and we'll raise a follow-up PR:

  • Yes please! Raise a follow-up PR: optimize the remaining 4 skills + add the Tessl skill-review-and-optimize GitHub Action
  • No thanks — happy with the 4 skills in this PR

Honest disclosure — I work at @tesslio where we build tooling around skills like these. Not a pitch - just saw room for improvement and wanted to contribute.

Want to self-improve your skills? Just point your agent (Claude Code, Codex, etc.) at this Tessl guide and ask it to optimize your skill. Ping me - @rohan-tessl - if you hit any snags.

Thanks in advance 🙏

Hey @m13v 👋

I ran your skills through `tessl skill review` at work and found some targeted improvements. Here's the full before/after:

| Skill | Before | After | Change |
|-------|--------|-------|--------|
| python-cdp-scripts | 10% | 90% | +80% |
| client-website | 80% | 93% | +13% |
| posthog-funnel | 81% | 85% | +4% |
| browser-script | 89% | 93% | +4% |

Changes:
- python-cdp-scripts: Added missing YAML frontmatter (name, description, trigger terms)
- client-website: Split 1307-line monolith into SKILL.md + references/ (COMPONENTS.md, DESIGN-SYSTEM.md)
- posthog-funnel: Improved description with action verbs, added error handling section
- browser-script: Improved description specificity with concrete actions
@rohan-tessl
rohan-tessl marked this pull request as ready for review April 13, 2026 10:18
m13v added a commit that referenced this pull request Jul 6, 2026
)

Root cause of lost drafts: a producer cycle that dies after its worker writes the
result but before consuming it strands the result in claude-queue/result/ (a
consumed result is deleted, so a surviving one = never consumed) -> no plan, no
merge, no card. Nothing detected or recovered these.

#1 Salvage (scripts/salvage_orphaned_prep_results.py, wired into run-draft-and-publish.sh):
   scans result/ for orphaned twitter-prep results (past the queue timeout so no
   live producer is racing, and within a max-age so stale threads are retired) and
   merges their candidates into the review queue via merge_review_queue.py (dedup-safe).
   Recovered 11 real stranded drafts on first run.

#2 Success/orphan visibility (claude_job.py): log a "consumed result for job X"
   event on the success path (the silent gap that made an orphaned result look
   identical to a healthy one); the salvage script logs the "ORPHAN ..." event.

#3 Correlation (claude_job.py): every provider.log line now carries batch=<BATCH_ID>
   so it links to the cycle's own twitter-cycle-<batch>.log.

#4 Remote visibility (scripts/relay_provider_log.py, wired into mcp/src/index.ts):
   tail claude-queue/provider.log to Cloud Logging (context="queue-provider"),
   incremental + self-locking + forward-only baseline, so a stranded batch on a
   customer box is queryable remotely instead of only on the box.

Also: refresh the CLAUDE.md note now that the persona lane is arm-aware.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant