ank — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited ank (Plugin) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
These are the skills I've formalised to make my own agentic workflow smoother and more efficient. Small, focused, composable — meant to slot into whatever I'm building rather than own the process. If any of them are useful to you too, take them, fork them, change them.
Pick one of the two install paths:
Claude Code plugin (native):
# In Claude Code:
/plugin marketplace add ananthanandanan/skills
/plugin install ank@ank-skillsSkills then appear namespaced as /ank:<skill> (e.g. /ank:review-board, /ank:handoff).
`npx skills` installer:
npx skills@latest add ananthanandanan/skillsThen pick the skills you want from the interactive list.
I push new skills and revise existing ones. Neither install path auto-updates — you have to pull changes in.
Plugin install — refresh the marketplace catalog, then update the plugin:
# In Claude Code:
/plugin marketplace update ank-skills
/plugin update ank@ank-skillsThe first command re-reads this repo's catalog so newly added skills are visible; the second pulls the updated skill files. Running /plugin on its own opens the manager UI if you'd rather browse than type. New skills show up as /ank:<skill> right away — restart Claude Code if one doesn't appear.
`npx skills` install — re-run the installer and reselect:
npx skills@latest add ananthanandanan/skillsThis overwrites the skill directories it manages, so local edits you've made to those files will be lost. Fork instead if you're customising.
/review-board — generates a single-file HTML "story-paced walkthrough" of a finished feature for human code review. Sidebar nav, Why/Tradeoff/Deferred callouts, status pills, reviewer's checklist. Use it after finishing a feature or merging a PR, when a teammate needs to understand the why and the shape of the change, not just the diff./handoff — drops a Markdown handoff doc at docs/<slug>-handoff.md distilling the current conversation (agenda, files touched, bugs, fixes, decisions, current state, next steps) so another agent — Claude, Codex, opencode, pi, ampcode, Cursor, or a fresh session — can resume without re-deriving context. Invoke when your context window is nearly full./third-degree — interviews you one question at a time, walking the decision tree until the spec is sharp. Each question comes with structured options and a recommended pick; anything the codebase already answers is inferred rather than asked. Front of the pipeline — invoke when the idea is fuzzy, before /spec or /visualise-plan./spec — turns a rough product idea (or a /third-degree summary) into a single opinionated SPEC.md at repo root (or docs/spec/<slug>.md for feature scope). One required-section list covers both library/SDK shapes and SaaS/app shapes; conditional sections (data model, API endpoints, frontend, pricing, success metrics) kick in based on what the product actually needs. Hard-stops until you reply "approved" — then /tasks picks up./tasks — turns an approved SPEC.md into TASKS.md (phases, parallel tracks, sized tasks, dependency map), then creates matching GitHub issues, a Project board, and empty-directory scaffolding from the spec's package layout. Confirms before any gh side-effect; degrades gracefully when gh is missing — it writes TASKS.md and asks you to gh auth login then re-run./visualise-plan — produces a single-file HTML implementation plan at docs/plan/<slug>.html (Botanical Almanac styling: milestones, data-flow SVG, mockups, files-touched list, key code). Use for feature-level planning when you want a visual artifact to review before writing code. Hard-stops until you reply "approved"./no-yap — answers in ASD-STE100 Simplified Technical English (the numbered-clause, one-idea-per-sentence style of an aircraft maintenance manual) instead of a wall of prose. Numbered sections with functional titles, ~20-word sentences, active voice, one word/one meaning, RESULT: and WARNING: on their own lines. Domain terms of art are never simplified — it compresses the language, not the subject — so it works as well on an enzyme pathway or a hydraulic failure as on a race condition. Chat-only, no file written; the style sticks for follow-ups until you ask for normal prose./expert-writing — rewrites a draft you already have (blog post, Slack message, architecture idea, README) into tighter, more technical writing in your own voice. Picks list vs paragraph per block based on the scenario, strips em dashes, curly quotes, and AI-slop vocabulary, and treats "more technical" as more substance (mechanism, trade-offs, failure modes) rather than more words. Gaps it can't ground get a visible [TODO] instead of an invented number. Returns the rewrite in chat, or edits the file in place when given a path.Add a directory under skills/<category>/<name>/ with a SKILL.md (YAML frontmatter must include name and description — CI enforces this), then append the path to the skills array in .claude-plugin/plugin.json. The review-board skill is the reference layout.
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.