refactor: extract non-critical reference detail to create skill headroom#95
Draft
saurabhrb wants to merge 1 commit into
Draft
refactor: extract non-critical reference detail to create skill headroom#95saurabhrb wants to merge 1 commit into
saurabhrb wants to merge 1 commit into
Conversation
Move lookup-only content out of three at-capacity skill bodies into references/ so ERP support can be added without breaching the 5000-token body cap. Behavior-neutral: only reference-lookup material moved; all decision logic and safety rules stay inline. - dv-connect: --validate output interpretation -> mcp-configuration.md (keep the 'not a success gate' rule inline) - dv-admin: pac data retention command syntax + arg table -> references/retention.md (keep agentic flow + retention-vs-bulk-delete decision inline) - dv-overview: Windows scripting rules -> references/windows-scripting.md Headroom: dv-connect 4->253, dv-admin 39->217, dv-overview 136->345. Static checks + version bump check pass. PATCH bump 1.6.0->1.6.1.
|
Azure Pipelines: Successfully started running 1 pipeline(s). Some pipeline(s) encountered errors during trigger evaluation. |
saurabhrb
marked this pull request as draft
July 14, 2026 15:14
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
Behavior-neutral refactor that frees token headroom in the three at-capacity skill bodies so that ERP (Finance & Operations) support can be added in a follow-up PR without breaching the 5000-token
SKILL.mdbody cap enforced byEVAL-BUDGET-02.No new capability, no routing change, no auth change — this is a pure Level-2 → Level-3 extraction following Anthropic's progressive-disclosure guidance and the repo's own
CLAUDE.mdtoken-budget rules.Why
Measured body token counts (tiktoken
cl100k_base, the counter the CI eval uses) onmainleft almost no room to add content:Extraction discipline
Only reference-lookup material was moved. All decision logic and safety rules stay inline in the body, because small-model agents sometimes read a reference link's filename and stop without opening it — so anything the agent needs to decide correctly or act safely must remain in Level 2.
--validatetwo-endpoint output interpretation intoreferences/mcp-configuration.md. The critical rule ("do NOT use--validateas a first-time success gate") stays inline.pac data retentioncommand syntax + argument table into a newreferences/retention.md. The agentic flow and the retention-vs-bulk-delete decision table stay inline. No safety-callout content touched.references/windows-scripting.md.Validation
python .github/evals/static_checks.py→ PASSED (8 skill files, 45 Python blocks, 10 categories)python .github/evals/version_bump_check.py→ PASSED (1.6.0 → 1.6.1, patch bump)Follow-up
The ERP support content (from the earlier ERP PRs) will land in a separate PR on top of this base, using thin gated routers inline + the ERP how-to detail in
references/erp-*.md.