v0.7.0 — context all-rounder + cross-run insight

The context all-rounder + cross-run insight. This release adds a layer of context-engineering tooling that keeps the agent's window lean on long, autonomous work, plus two read-only views into how your runs are going.

Highlights

• /ss-context — a standing-context budget cockpit. It measures your always-loaded footprint (CLAUDE.md, skill descriptions, STATE.md / CONTEXT.md) against a budget, verdicts OK / WARN / OVER, and warns you before you blow it — automatically, at session start. (context all-rounder, Front 1)

• ss-ctx runtime-output sandbox — two halves over one store:

  • a PostToolUse hook that transparently shrinks oversized clean Bash output, offloading the full text to a retrievable store;
  • a dependency-free Node MCP server exposing ctx_execute / ctx_batch_execute (run a command, keep its verbose output out of context), ctx_search / ctx_show (over the shared store), and ctx_fetch_and_index (fetch a URL, store the text, return a preview).

  So noisy command runs and fetched pages never flood the context window. (Front 2)

• /ss-stats — DORA-style cross-run analytics: a per-run table plus a rollup (gate-fail rate, skips, and an improving / worsening / flat trend).

• /ss-trace — a change's full provenance, joining its spec/plan docs, ledger gate events, and git commits into one chronological lineage.

30 skills · 1 MCP server. Everything ships bash + PowerShell twins (the MCP server is a single cross-platform Node process) and is covered by the self-test suite + CI.

Full detail in the Changelog. MIT licensed.

v0.6.0

The verification track — two read-only commands that prove a project is set up right and that the build stayed on plan. 26 skills total.

/ss-doctor — project health check (the verify leg)

The diagnostic counterpart to /ss-init: runs five checks (jq, git, .superstack/config, gitignore, ledger) and prints a [OK]/[WARN]/[FAIL] checklist with an actionable fix on every non-OK line. Validates the ledger jq-free, so it works even while reporting jq missing. Exits 0 (healthy/warnings) or 1 (problems) — usable as a CI preflight. bash + PowerShell.

/ss-drift — plan-vs-build drift detection

Compares a plan's declared **Files:** against what the branch actually changed (base...HEAD + working tree + untracked), reporting unplanned changes (scope creep) and planned-but-untouched files. Desired-vs-actual drift, after Terraform's model, applied to "did the implementation stay within the approved plan." Exits 1 on drift for CI. bash + PowerShell.

Full changelog: https://github.com/Mrshahidali420/superstack/blob/main/CHANGELOG.md

v0.5.0

Two additions to the SuperStack loop (24 skills total).

/ss-init — per-project bootstrap

Make a project loop-ready in one idempotent, non-destructive command: writes a default .superstack/config (with the tunable mandatory_phases / evolve_threshold keys), ensures .superstack/ is gitignored, and records a genesis ledger entry. --dry-run previews the plan; --force resets the config. The per-project counterpart to the global install.sh. bash + PowerShell.

/ss-replay — the "story" leg of the proof trio

Replays a loop run from the ledger as a chronological ASCII timeline (elapsed time, phase, event, PASS/FAIL/SKIP, (retry) tags) with a footer of story stats. --save writes a shareable fenced Markdown file to .superstack/replays/. Completes the proof trio: audit (gate) + report (stats) + replay (story). bash + PowerShell.

Full changelog: https://github.com/Mrshahidali420/superstack/blob/main/CHANGELOG.md

v0.4.0 — ss-evolve --since + --explore

SuperStack v0.4.0 extends /ss-evolve with two new flags — both cross-platform (bash + PowerShell), byte-identical output, fully tested.

Added

• ss-evolve --since <window> (Nd / Nh / YYYY-MM-DD) restricts detection to a recent slice of the ledger; composes with every other flag.
• ss-evolve --explore scaffolds a draft new-skill proposal into .superstack/proposals/<name>/ (Tier 2) — never committed; the /ss-evolve skill authors the body and a human promotes it. Dedups independently of --apply via .superstack/explore-state.

Changed

• Repositioned the README, CLAUDE.md, and the plugin/marketplace description around SuperStack as its own framework — added why-it's-better, a capability comparison, use cases, and benefits; reframed credits as inspiration rather than a "distillation."

Full changelog: https://github.com/Mrshahidali420/superstack/blob/main/CHANGELOG.md

v0.3.0 — /ss-report + /ss-evolve

Acting on the ledger

v0.2.0 gave SuperStack a Loop Ledger and a proof-of-process gate. v0.3.0 puts that signal to work — two features that read the ledger.

Added

• /ss-report — a shareable, copy-pasteable Markdown summary of how a change was built (phases run, skip reasons, elapsed time, change size), sourced from the ledger + git. Drop it in a PR or a status update. Read-only; never gates. scripts/ss-report (+ PowerShell twin); --save writes .superstack/run-report-<change>.md.
• /ss-evolve — project self-improvement from the ledger. It detects recurring patterns (a phase skipped >= a threshold, a gate that repeatedly fails) and auto-applies low-risk CONTEXT.md insights as revertable chore(evolve): commits, while routing brand-new skill drafts to .superstack/proposals/ for your review (never auto-committed). scripts/ss-evolve (+ PowerShell twin); --dry-run previews; deduped via .superstack/evolve-state; threshold via .superstack/config evolve_threshold (default 3).

Both ship bash + PowerShell twins with byte-identical output and are covered by the self-test (tests/run.sh is now [1/6]..[6/6]). 22 skills, 4 agents, 2 hooks.

Install: /plugin marketplace add Mrshahidali420/superstack then /plugin install superstack@superstack

Full diff: v0.2.0...v0.3.0

v0.2.0 — Loop Ledger

Loop Ledger — proof of process

SuperStack now records its loop's execution to .superstack/ledger.jsonl and can verify the loop actually ran before you ship.

Added

• Loop Ledger: a ledger helper (bash + PowerShell) plus the /ss-audit proof-of-process gate — it checks that every mandatory phase (default review,secure) cleared its gate or carries an explicit skip-with-reason. /ss-ship runs it and attaches a SuperStack process: … attestation to the PR. An opt-in enforcement hook (SUPERSTACK_AUDIT=1) blocks pushes when the loop is incomplete.
• Eight supporting skills — /ss-debug, /ss-guard, /ss-respond, /ss-worktree, /ss-pause, /ss-resume, /ss-retro, /ss-docs — plus /ss-audit (20 skills total).
• Cross-agent install: install.sh --host codex|cursor|opencode|factory|kiro|--all (and the PowerShell equivalent).
• Ralph loop: --dry-run preview, per-iteration run logs, archive-on-completion.
• Hooks: a SessionStart bootstrap (cross-platform polyglot launcher) and an opt-in PreToolUse guard.
• CHANGELOG.md, tag-triggered CI, and SPDX headers on the scripts.

Changed

• Linter upgraded from a structural check to a quality check (trigger-style descriptions, a single H1, resolvable [[wikilinks]], loop completeness).

Install: /plugin marketplace add Mrshahidali420/superstack then /plugin install superstack@superstack

Full diff: v0.1.0...v0.2.0

SuperStack v0.1.0

One disciplined loop for coding agents — Frame -> Plan -> Build -> Review -> QA -> Secure -> Ship -> Learn — distilled from Superpowers, GSD, gstack, and Ralph, with Karpathy's four anti-mistake laws baked in.

Install

/plugin marketplace add Mrshahidali420/superstack
/plugin install superstack@superstack

Verified via claude plugin install: the marketplace install discovers 11 skills + 4 agents (~827 always-on tokens/session). Marketplace version-pinning uses the superstack--v0.1.0 tag.

What's inside

• 9 phase commands (/ss-frame … /ss-ship) + /ss-ralph autonomy + /ss-help index
• superstack bootstrap skill that auto-loads the methodology on install
• 4 subagents: planner, code-reviewer, security-reviewer, qa-runner
• Ralph autonomous loop (bash + PowerShell), cross-platform installers, plugin + marketplace manifests
• CI lint-check (skill frontmatter + JSON) on every push/PR

Notes

Original distillation — credits Superpowers, GSD, gstack, Ralph, and Karpathy; vendors none of them. See CREDITS.md. MIT licensed.