Summary
Several skill files are large enough that agents must load a lot of context before reaching the specific workflow they need. This is especially visible in skills/insforge-cli/SKILL.md, which already has a references directory but still embeds many detailed workflow sections inline.
For a skills repo, repo cleanup should include a context hygiene pass: keep entrypoint SKILL.md files as routing/triage documents and move deep workflow bodies into focused references.
Validation
Valid. A local file-size/line-count scan on main@fe033166dd578efe2d0f59f3f9a950123cd22fda found:
skills/insforge-cli/SKILL.md - 687 lines / 41562 bytes
skills/insforge-integrations/references/okx-x402.md - 990 lines / 39127 bytes
skills/insforge-integrations/references/better-auth.md - 648 lines / 33170 bytes
skills/insforge/auth/sdk-integration.md - 487 lines / 18320 bytes
The main CLI skill already links many references but still includes long sections for migrations, deployments, compute, schedules, diagnostics, CI, and .insforge/project.json.
Relevant Code
skills/insforge-cli/SKILL.md:70 through skills/insforge-cli/SKILL.md:275 - command reference overview.
skills/insforge-cli/SKILL.md:328 through skills/insforge-cli/SKILL.md:364 - detailed migration workflow repeated in the main skill body.
skills/insforge-cli/SKILL.md:375 through skills/insforge-cli/SKILL.md:430 - detailed frontend deployment workflow.
skills/insforge-cli/SKILL.md:435 through skills/insforge-cli/SKILL.md:489 - detailed compute workflow.
skills/insforge-cli/SKILL.md:507 through skills/insforge-cli/SKILL.md:608 - detailed schedules workflow.
Proposed Solution
Do a progressive-disclosure pass:
- Keep each
SKILL.md focused on trigger/routing, critical rules, and reference links.
- Move long workflow bodies into existing or new
references/*.md files.
- Add a small "load only the reference you need" pattern to the entrypoints.
- Preserve critical warnings in the entrypoint when they affect safety.
Acceptance Criteria
- Entrypoint skill files are meaningfully shorter and faster to scan.
- No workflow content is lost; detailed docs live in focused references.
- Cross-links remain valid after the split.
- The Markdown link/frontmatter validator from the DX issue passes after the restructure.
Summary
Several skill files are large enough that agents must load a lot of context before reaching the specific workflow they need. This is especially visible in
skills/insforge-cli/SKILL.md, which already has a references directory but still embeds many detailed workflow sections inline.For a skills repo, repo cleanup should include a context hygiene pass: keep entrypoint
SKILL.mdfiles as routing/triage documents and move deep workflow bodies into focused references.Validation
Valid. A local file-size/line-count scan on
main@fe033166dd578efe2d0f59f3f9a950123cd22fdafound:skills/insforge-cli/SKILL.md- 687 lines / 41562 bytesskills/insforge-integrations/references/okx-x402.md- 990 lines / 39127 bytesskills/insforge-integrations/references/better-auth.md- 648 lines / 33170 bytesskills/insforge/auth/sdk-integration.md- 487 lines / 18320 bytesThe main CLI skill already links many references but still includes long sections for migrations, deployments, compute, schedules, diagnostics, CI, and
.insforge/project.json.Relevant Code
skills/insforge-cli/SKILL.md:70throughskills/insforge-cli/SKILL.md:275- command reference overview.skills/insforge-cli/SKILL.md:328throughskills/insforge-cli/SKILL.md:364- detailed migration workflow repeated in the main skill body.skills/insforge-cli/SKILL.md:375throughskills/insforge-cli/SKILL.md:430- detailed frontend deployment workflow.skills/insforge-cli/SKILL.md:435throughskills/insforge-cli/SKILL.md:489- detailed compute workflow.skills/insforge-cli/SKILL.md:507throughskills/insforge-cli/SKILL.md:608- detailed schedules workflow.Proposed Solution
Do a progressive-disclosure pass:
SKILL.mdfocused on trigger/routing, critical rules, and reference links.references/*.mdfiles.Acceptance Criteria