Problem
The Tessl skill review workflow fails on every pull request with the current reviewer, including PRs that do not touch the runtime skill. On PR #52 (docs only, git diff v1.2.0 -- skills empty) it scored the released skills/java-streams/SKILL.md at 91%, below the workflow's threshold of 100:
Content: 79% (conciseness 4/5, actionability 4/5, workflow_clarity 4/5, progressive_disclosure 5/5)
Description: 100%
Review Score: 91%
Run: https://github.com/martinfrancois/java-streams-skill/actions/runs/35521420116/job/106106052899
The file has not changed since v1.2.0 scored 100. Tessl replaced the single-pass skill review with Tessl Review (an agent over the whole bundle) and documents that scores can differ: https://docs.tessl.io/improving-your-skills/migrate-from-skill-review
Reviewer suggestions
- Tighten the opening preservation paragraph and the find-first rule prose; move directive sentences into the existing decision tables.
- Add a second complete executable example (a
groupingBy or toMap snippet with merge and null handling) next to the partial flatMap one-liner.
- Surface a validation checkpoint earlier in the workflow (after step 1's API selection) rather than only in step 7.
Why this is not fixed in #52
Any edit to SKILL.md changes the measured runtime bundle. Per docs/agents/workflow.md that requires the quality review at 100 plus targeted, main, reference (both variants), and regression (with context) reruns before release-readiness can be claimed: about 390 Tessl credits on the current free plan. #52 is a docs-only companion pointer and must not carry that.
Proposed fix
- Apply the three suggestions in a
fix(skill) PR, keeping every rule's meaning.
tessl review run --workspace martinfrancois --threshold 100 skills/java-streams/SKILL.md until 100.
- Run the full hosted sequence from
docs/agents/pre-submit-gate.md and the composition check from docs/agents/ownership-boundaries.md (scripts/run_composed_eval.sh in java-functional-style-skill).
- Release through Release Please; confirm the registry score stays at or above 2.22x.
The check is not in the ruleset's required list, so it does not block merging, but it shows red on every PR until this lands.
AI Assistance (if used)
Problem
The
Tessl skill reviewworkflow fails on every pull request with the current reviewer, including PRs that do not touch the runtime skill. On PR #52 (docs only,git diff v1.2.0 -- skillsempty) it scored the releasedskills/java-streams/SKILL.mdat 91%, below the workflow's threshold of 100:Run: https://github.com/martinfrancois/java-streams-skill/actions/runs/35521420116/job/106106052899
The file has not changed since v1.2.0 scored 100. Tessl replaced the single-pass skill review with Tessl Review (an agent over the whole bundle) and documents that scores can differ: https://docs.tessl.io/improving-your-skills/migrate-from-skill-review
Reviewer suggestions
groupingByortoMapsnippet with merge and null handling) next to the partialflatMapone-liner.Why this is not fixed in #52
Any edit to
SKILL.mdchanges the measured runtime bundle. Perdocs/agents/workflow.mdthat requires the quality review at 100 plus targeted, main, reference (both variants), and regression (with context) reruns before release-readiness can be claimed: about 390 Tessl credits on the current free plan. #52 is a docs-only companion pointer and must not carry that.Proposed fix
fix(skill)PR, keeping every rule's meaning.tessl review run --workspace martinfrancois --threshold 100 skills/java-streams/SKILL.mduntil 100.docs/agents/pre-submit-gate.mdand the composition check fromdocs/agents/ownership-boundaries.md(scripts/run_composed_eval.shin java-functional-style-skill).The check is not in the ruleset's required list, so it does not block merging, but it shows red on every PR until this lands.
AI Assistance (if used)