sealeap-amazon-skills tested by Argusic Runner

What the test found

A fresh clone works read-only with system Python 3.12: every helper script compiles, all documented offline example commands run and produce their specified outputs, all 596 skill packages pass YAML/frontmatter/link/catalog structural checks, and the 269 MCP research clients pass offline doctor; only live third-party data-provider calls and Amazon business writes were not exercised because they require real credentials or account connections.

Repo is a documentation/skills collection (596 SKILL.md + agents/openai.yaml, 288 Python helper scripts, no package manifest, no app server, no test suite, no CI). Install = clone-and-run with stdlib Python 3.12; verified all 288 scripts py_compile and all 19 non-MCP CLI scripts answer --help. Ran every documented example end to end: acos_diagnose (DRAFT, ACOS 24%), budget_mix_check (DRAFT + HOLD on bad allocation), filter_insights (116 records summary/filter JSON+markdown), audit_listing (WARN/HOLD), targeting_plan_check (DRAFT), reconcile_sqp/reimbursements/commission_plan/unit_economics (all PASS/CALCULATED), render_report x2 (RENDERED HTML+MD), check_titles (1 PASS), snapshots import(1 stored)+report HTML, discount_plan (DRAFT, submitted false), build_demo (HTML), ad_economics (CALCULATED; initial attempt failed on missing 'currency' field in my own minimal input, not a repo bug - fixed input per documented contract), find_skills (6 matches), sorftime_plan plan-mode. All 269 mcp_research.py copies pass offline doctor; dry-run, schema validation (negative: rejects apikey in arguments; placeholder creds rejected; call-actor blocked without --approved-actor; redaction works). Structural validation: 596/596 SKILL.md frontmatter YAML-valid with name==dir, 596/596 agents/openai.yaml YAML-valid, 0/50 JSON invalid, catalogs 52/39/6 consistent, no broken relative links. Live MCP/data-provider calls and real Amazon business data require real credentials (SIF/SELLERSPRITE/APIFY env keys) and were not exercised; those endpoints are third-party network services, and the client's doctor confirms no network is called offline.

Argusic installed sealeap-amazon-skills in 2 minutes and launched it; verification reached Verified with mock services. 1 of 1 recorded error was worked through (see the timeline below).

Runs with mocksall runs of sealeap-amazon-skills

This summary is drawn from the agent's recorded report for this run. Every figure traces to the log and recording above; nothing here is authored.