Implement the tone refinement skill #2

Closed
opened 2026-06-04 08:05:52 +00:00 by tsieprawski · 12 comments
Owner

Plan:

  1. Define the skill scope and audience.
  2. Translate the bass/guitar tone-refinement workflow into reusable skill instructions.
  3. Capture the required intake questions and sample-capture requirements.
  4. Encode the core refinement method:
    • real-world routing constraints first
    • fresh target-specific sample capture
    • per-string and per-articulation evaluation
    • split-path design when useful and feasible
    • IR and gain-staging balance before final loudness trim
  5. Add skill structure and repository content:
    • skill manifest/instructions
    • README with purpose and usage
    • example workflow or checklist
  6. Validate the skill for clarity, public readability, and practical reuse.

Implementation notes:

  • The skill should work for bass and guitar, but bass split-path workflows are the initial reference case.
  • It must avoid impossible routings or hardware/software assumptions.
  • It should prefer reproducible A/B steps over vague tone advice.
  • It should tell the user to create fresh samples for the actual instrument, tuning, settings, and target use-case.
Plan: 1. Define the skill scope and audience. 2. Translate the bass/guitar tone-refinement workflow into reusable skill instructions. 3. Capture the required intake questions and sample-capture requirements. 4. Encode the core refinement method: - real-world routing constraints first - fresh target-specific sample capture - per-string and per-articulation evaluation - split-path design when useful and feasible - IR and gain-staging balance before final loudness trim 5. Add skill structure and repository content: - skill manifest/instructions - README with purpose and usage - example workflow or checklist 6. Validate the skill for clarity, public readability, and practical reuse. Implementation notes: - The skill should work for bass and guitar, but bass split-path workflows are the initial reference case. - It must avoid impossible routings or hardware/software assumptions. - It should prefer reproducible A/B steps over vague tone advice. - It should tell the user to create fresh samples for the actual instrument, tuning, settings, and target use-case.
Author
Owner

Implementation started on branch impl-skill. Planned repo shape: root SKILL.md, agents/openai.yaml, and a small references/ set for sample capture and evaluation workflow.

Implementation started on branch impl-skill. Planned repo shape: root SKILL.md, agents/openai.yaml, and a small references/ set for sample capture and evaluation workflow.
tsieprawski referenced this issue from a commit 2026-06-04 08:11:23 +00:00
Author
Owner

Initial implementation is on branch impl-skill and in PR !3. Current scope: root SKILL.md, agents/openai.yaml, capture/evaluation reference, and updated REUSE metadata. Validation passed with quick_validate.py and reuse lint.

Initial implementation is on branch impl-skill and in PR !3. Current scope: root SKILL.md, agents/openai.yaml, capture/evaluation reference, and updated REUSE metadata. Validation passed with quick_validate.py and reuse lint.
Author
Owner

Added another implementation pass on impl-skill and PR !3: fresh-strings guidance, user-run DAW measurement workflow, common-vs-dedicated preset intake, stronger final mix validation, and a separate IR-selection reference.

Added another implementation pass on impl-skill and PR !3: fresh-strings guidance, user-run DAW measurement workflow, common-vs-dedicated preset intake, stronger final mix validation, and a separate IR-selection reference.
Author
Owner

Another pass added to PR !3: IR-selection now covers open-ended multi-label categorization, naive yes/no first-pass filtering, pass templates for note-taking, simple single-track switching in early phases, optional bandmate A/B input, low-fatigue bracket rules, and keeping first- and second-choice winners per label.

Another pass added to PR !3: IR-selection now covers open-ended multi-label categorization, naive yes/no first-pass filtering, pass templates for note-taking, simple single-track switching in early phases, optional bandmate A/B input, low-fatigue bracket rules, and keeping first- and second-choice winners per label.
tsieprawski referenced this issue from a commit 2026-06-04 08:58:57 +00:00
Author
Owner

Latest PR !3 pass: IR-selection now notes the process may take 2-6 hours split over evenings, mentions workshop-style group use, moves note-taking templates into references/ir-selection-templates.txt, and adds quick-write league and tournament templates.

Latest PR !3 pass: IR-selection now notes the process may take 2-6 hours split over evenings, mentions workshop-style group use, moves note-taking templates into references/ir-selection-templates.txt, and adds quick-write league and tournament templates.
Author
Owner

Latest PR !3 pass: added references/bad-tone-patterns.md and linked it from SKILL.md. It currently includes two real diagnosis patterns: single-output clanky drive that loses low end in mix, and over-squashed compression that evens strings but makes the tone too flat.

Latest PR !3 pass: added references/bad-tone-patterns.md and linked it from SKILL.md. It currently includes two real diagnosis patterns: single-output clanky drive that loses low end in mix, and over-squashed compression that evens strings but makes the tone too flat.
Author
Owner

Latest PR !3 pass: added another bad-result pattern for multi-path rigs forced to collapse to one output because of downstream splitter/FOH/IEM limits, and tightened SKILL.md so external output infrastructure is treated as part of the routing constraint set.

Latest PR !3 pass: added another bad-result pattern for multi-path rigs forced to collapse to one output because of downstream splitter/FOH/IEM limits, and tightened SKILL.md so external output infrastructure is treated as part of the routing constraint set.
Author
Owner

Latest PR !3 pass: added references/platform-tradeoffs.md and linked it from SKILL.md. It covers DAW rig vs digital modeler vs pedalboard/amp/cab tradeoffs, including automation/routing advantages, live logistics benefits, and the practical risks of laptop rigs.

Latest PR !3 pass: added references/platform-tradeoffs.md and linked it from SKILL.md. It covers DAW rig vs digital modeler vs pedalboard/amp/cab tradeoffs, including automation/routing advantages, live logistics benefits, and the practical risks of laptop rigs.
Author
Owner

Latest PR !3 pass: expanded platform-tradeoffs with the extra staffing and rack-level complexity of more-digital rigs, plus the upside of integrated samples and cheap live recording for later mix refinement and live review.

Latest PR !3 pass: expanded platform-tradeoffs with the extra staffing and rack-level complexity of more-digital rigs, plus the upside of integrated samples and cheap live recording for later mix refinement and live review.
tsieprawski referenced this issue from a commit 2026-06-04 09:32:17 +00:00
Author
Owner

Latest PR !3 pass: added check.sh plus .forgejo/workflows/check.yaml for pull_request CI. The workflow uses the same bare git clone/fetch/checkout style as services-gator instead of checkout actions, and local check.sh is green.

Latest PR !3 pass: added check.sh plus .forgejo/workflows/check.yaml for pull_request CI. The workflow uses the same bare git clone/fetch/checkout style as services-gator instead of checkout actions, and local check.sh is green.
Author
Owner

Latest PR !3 pass: expanded check.sh and Forgejo PR CI to cover shellcheck, shfmt, yamllint, and typos, with .yamllint.yaml and .typos.toml added to keep the checks practical for this repo. Local check.sh is green.

Latest PR !3 pass: expanded check.sh and Forgejo PR CI to cover shellcheck, shfmt, yamllint, and typos, with .yamllint.yaml and .typos.toml added to keep the checks practical for this repo. Local check.sh is green.
Author
Owner

Latest PR !3 pass: fixed the Alpine CI failure caused by PEP 668 / externally managed system Python. The workflow now installs reuse inside a temporary virtualenv instead of pip-installing into the system environment.

Latest PR !3 pass: fixed the Alpine CI failure caused by PEP 668 / externally managed system Python. The workflow now installs reuse inside a temporary virtualenv instead of pip-installing into the system environment.
tsieprawski referenced this issue from a commit 2026-06-04 09:47:42 +00:00
Sign in to join this conversation.
No labels
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
tsieprawski/tone-agent-skill#2
No description provided.