Skill Optimizer

Iteratively review and optimize local Tessl skills using tessl skill review and tessl skill review --optimize, tracking improvement across iterations.

Workflow

1. Discover skills
2. Baseline review each skill
3. Optimize and re-review in a loop
4. Present results

Step 1: Discover Skills

find . -name "SKILL.md" -not -path "./.tessl/*" -not -path "*/node_modules/*"

If no skills are found, ask the user to provide the path to their skills directory. If multiple skills are found, present the list and ask which to optimize (default: all).

Step 2: Baseline Review

For each selected skill, run a baseline review (<path-to-skill-directory> is the parent directory of SKILL.md):

tessl skill review <path-to-skill-directory> --json

Record overall score, individual category scores, and any warnings or failures. Present a summary table before proceeding:

Baseline Review Results
=======================
Skill                          Score
─────────────────────────────  ─────
running-dbt-commands           72%
troubleshooting-dbt-job-errors 65%

Step 3: Optimize Loop

For each skill, iterate up to 3 cycles:

3a. Snapshot — Read and store the current SKILL.md content.

3b. Optimize

tessl skill review <path-to-skill-directory> --optimize

3c. Diff — Read the updated SKILL.md and diff against the snapshot. Record changes for the final report.

3d. Re-review

tessl skill review <path-to-skill-directory> --json

3e. Stop conditions

• Score ≥ 90%: stop — skill conforms to best practice.
• Score improved but < 90%: run another cycle.
• Score did not improve or decreased: stop — further iterations unlikely to help.
• 3 cycles reached: stop regardless.

Step 4: Present Results

See references/output-format.md for the baseline summary table and final results template.