Agent Skills by Gordon Shumway
Showing 16 vetted skills indexed across 1 GitHub repositories.
veegee82drift-detection
Scan codebases for cumulative drift across seven indicator categories.
veegee82e2e-driven-iteration
Run an E2E test before each fix and track loss convergence.
veegee82dialectical-reasoning
Enforce thesis, antithesis, and synthesis before non-trivial recommendations.
veegee82iterative-refinement
Refine completed deliverables by targeting observed defects in a second pass.
veegee82host-statusline
Render a permanent LDD progress statusline in Claude Code.
veegee82method-evolution
Detect recurring rubric violations and propose a single method change with measured loss.
veegee82dialectical-cot
Enforce thesis, antithesis, and synthesis with per-step calibration for verifiable reasoning.
veegee82define-metric
Register agent-defined metrics as loss components with calibration gates.
veegee82bootstrap-userspace
Detect host persistence surfaces and initialize the LDD trace store at session start.
veegee82reproducibility-first
Require reproducible evidence before editing code for CI failures or flaky tests.
veegee82architect-mode
Guide architecture design through a five-phase protocol with scoring rubric.
veegee82using-ldd
Route user intents to LDD sub-skills via trigger-phrase mappings.
veegee82loss-backprop-lens
Apply gradient-descent principles to code edits when patching failing tests.
veegee82docs-as-definition-of-done
Enforce documentation updates for public API, CLI flags, and config shapes before commit.
veegee82loop-driven-engineering
Coordinate four nested loops for engineering work within a five-iteration budget.
veegee82root-cause-by-layer
Diagnose bugs through a five-layer ladder from symptom to conceptual origin.