Harden portable pstack runtime neutrality

[Go7hic]

Summary

This converts the portable distribution from “Cursor text with substitutions” into a capability-first Agent Skills system with enforceable portability rules.

What changed

Portable regression gate

• added scripts/audit_portability.py
• added .github/workflows/portable-audit.yml
• validate skill frontmatter, unique names, required inventories, and mirror equality
• reject concrete Cursor model slugs, Cursor-only helper fields, fixed transcript paths, external plugin dependencies, and ambiguous mechanical-port wording
• run a repository-wide audit plus a strict changed-file audit on every pull request

Runtime-neutral workflow layer

• rewrote pstack and poteto-mode around the capability contract
• rewrote how, why, architect, swarm, reflect, recall, automate-me, and show-me-your-work
• removed fixed transcript-directory and session-schema assumptions
• changed model selection to model_role
• made mode lifetime and degraded-host behavior explicit
• kept lead-agent ownership of synthesis, final diff judgment, and verification

Portable playbooks

Updated both mirrored playbook trees for:

• Babysit
• Eval
• Hillclimb
• Opening a PR
• Perf issue
• Session pickup

These now use capabilities and active-host forge/session/runtime access rather than Cursor-specific workflows.

Maintenance and documentation

• documented canonical mirror sources and semantic review requirements
• made setup, cleanup, unattended-run, and skill-customization guides host-aware
• made port_pass2.py conservative and targeted
• clarified what the portable layer guarantees and what must degrade when a host lacks subagents, model selection, persistent modes, or runtime control

Validation

• maintenance scripts compile
• playbook, adapter, and capability-contract mirrors are byte-identical
• full repository audit: 0 error(s), 0 warning(s)
• strict changed-file audit: 0 error(s), 0 warning(s)
• GitHub Actions Portable audit completed successfully on head 2cbdc29112587664970a62c536648461589dabd1

Scope

This PR establishes a clean portable baseline and prevents regressions. Deeper work such as host conformance evals, an idempotent upstream-import pipeline, upstream provenance manifests, adapter contract versioning, and model-override schema validation remains tracked in #2.

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.
To continue using code reviews, you can upgrade your account or add credits to your account and enable them for code reviews in your settings.