igor1309
@igor1309
Public Skills
agister
by igor1309
Drive Agister from outside - find out which workflows and processes the settings in use define and what each takes, run a workflow in your worktree, continue a run, read a run's record and what it left in the worktree, and prepare a repository's .agister/. Use when asked to "use agister", to run work through Agister, to read what a run produced, or to set up .agister/ in a repository.
adversarial-cycle
by igor1309
Multi-agent execution engine that takes a committed brief to green, committed code — an orchestrator resolves a skill-owned runtime profile, launches a Planner, one Implementer per step, a cold Reviewer, and an independent review, holds cycle state in a committed tracker, and ends in done or stopped.
worktree
by igor1309
Create a new git worktree, sync shared config into it via the bundled sync script (never replicate config by hand or hand-author symlinks), and verify the resulting worktree is clean. Triggers on "create a worktree", "new worktree", "set up a worktree". Self-contained — the worktree creator and config-sync script ship with this skill under scripts/.
figma-use
by igor1309
"MANDATORY prerequisite — you MUST invoke this skill BEFORE every use_figma tool call. NEVER call use_figma directly without loading this skill first. Skipping it causes common, hard-to-debug failures. Trigger whenever the user wants to perform a write action or a unique read action that requires JavaScript execution in the Figma file context — e.g. create/edit/delete nodes, set up variables or tokens, build components and variants, modify auto-layout or fills, bind variables to properties, or inspect file structure programmatically."
doc-discipline
by igor1309
Use before drafting or restructuring an ADR, spec, grounding or research note, plan, tracker or log entry; before citing a commit in durable prose; before adding a link between two docs; and when retiring a plan or tracker that others link to. Triggers on "write the doc", "review the doc", "cold-reader pass", "should this doc link to", "cross-link", "link the docs", "retire the plan", "delete the tracker", "supersede the doc", "cite the commit", "reference the SHA", "log entry".
skill-conventions
by igor1309
Conventions for building, organizing, reviewing, and improving agent skills — the knowledge/rules/procedure taxonomy, the "hook enforces, skill explains, script executes" pattern, folder shape (scripts/references/backlog/adr), frontmatter gates, and a ten-axis quality review. Use when adding a new skill, auditing or improving an existing one, deciding whether a piece of automation belongs in a hook, a script, or prose, or refactoring a skill library. Triggers on "new skill", "skill conventions", "review skill", "is this skill effective", "improve the skill description", "skills audit", "where should this rule live".
postmortem
by igor1309
Reflect on the current session and propose polish/tweaks/fixes for a named skill, grounded in evidence from this conversation. Use when wrapping up a session that exercised a skill — to capture what worked, what should be fixed, and what adjacent skills were missing. Sibling to skill-conventions, but session-scoped instead of corpus-scoped.
strict-tdd
by igor1309
Strict TDD mode — RED/GREEN/REFACTOR discipline with no-junk scope lock
background-task-liveness
by igor1309
Use when supervising a long-running background process — a build, test run, CI pipeline, deploy, or data job — and you need to choose what to poll, how often, how to tell a slow run from a wedged one, and how to keep monitor state across a teardown. Triggers on "monitor the build", "is it still running", "the build is hanging", "watch the deploy", "poll CI", "arm a monitor", "wedged", "stalled", "no output for a while", "background task".
code-review-discipline
by igor1309
Use before declaring any implementation complete or reporting results (self-review is unprompted), when reviewing a diff or someone else's code, before writing down any fail/warning finding, and when turning review findings into an implementation plan. Triggers on "review the diff", "self-review", "review this code", "adversarial review", "before I call this done", "is this a finding", "fail/warning", "regression risk", "test seam", "plan from the findings".
pr-ci-watch
by igor1309
Use when the user wants to monitor GitHub PR CI status, wait for checks to finish, merge on green, or investigate failed checks. Triggers on phrases like "is CI done", "wait for checks", "watch the PR", "merge when green", or after pushing commits that invoke CI workflows. Prevents wasteful polling of gh pr checks and encodes the blocking/background decision.
intake
by igor1309
Use when starting work on a task — assigned implementation, bug report, vague requirement, or open-ended exploration — even if the user doesn't say "intake" or "research." Trigger phrases: "/intake", "here's your task", "implement this", "fix this bug", "there's a bug", "explore X", "before we commit", or a bare task reference (step number, PR, doc path). Planning or coding from shallow context produces wrong solutions; this skill forces a deliberate context-capture step before any downstream work begins. Stops cleanly — no auto-transition to planning or implementation.
debug-print
by igor1309
Isolate bugs by adding debugPrint logging only — no behavior changes. Use when the user wants to add debug prints to trace a bug. Argument is the bug description.
release-process
by igor1309
Per-unit release discipline — model, invariants, and verification gates
using-dev-skills
by igor1309
Use when starting any conversation - maps the dev-skills toolkit by ladder (TDD, spec, review, planning, release, Swift) so the right skill is invoked before any response or action.
agonara-coach
by igor1309
Tutor a student against an Agonara instance — the explanation/tutoring front of the student assistant. Explain a concept at the depth the student asks for, take it deeper on request, diagnose why a weak area is weak from the evidence behind it before re-teaching, teach prerequisites first when the base is shaky, and teach the discriminating feature of confusable concepts. Grounds every explanation in accepted materials, never in your own knowledge unstated. Owns explanation craft only; the API reads, the assessment-integrity boundary, and the MCQ item-quality bar stay with agonara-student, practice with the drill skills. Use when the task is to tutor — explain or help a student understand a topic — rather than to take a check or do API work.
agonara-student
by igor1309
Assist a human Agonara Student through the instance's HTTP API — discover and preview programs without creating state, self-join/leave/rejoin and act on invitations, resume learning and follow the next step, navigate accepted concepts and materials and read hosted material content where present, present checks and relay the student's own quick, open, and readiness answers, and read feedback, results, and the evidence behind progress. Owns the API, the assessment-integrity boundary (an agent may help a student learn but must never produce, choose, or submit the student's assessment evidence), and the MCQ item-quality bar — the craft of explaining a topic and diagnosing why an area is weak is owned by agonara-coach. The student sees only accepted corpus and its own state — never curator drafts, answer keys, correctness metadata, or evaluator internals. Use when an agent acts as a student's assistant against an Agonara instance — the API, boundary, and checks substrate the coach and drill skills build on.
button-quiz-loop
by igor1309
Use when running a button-based quiz/drill in chat and continuity matters more than commentary.
agonara-curator-corpus
by igor1309
Assist a human Agonara Program Curator — do the corpus API work on their behalf to assemble program structure, modeling what the curator hands you (a section, an idea, a document to host) into the right corpus shape and asking when the shape is unclear — create a program, propose topics and concepts as corpus candidates, add material references and relationships through navigation candidates, host readable material content for students, keep drafts separate from accepted corpus, explicitly accept or reject candidates, read accepted corpus and revisions, and read corpus events; treat your own drafts and reports as untrusted until the curator verifies each candidate against trusted state, accept per entity rather than in a batch, and declare contamination on any procedural violation. Use when a curator's assistant agent is building or refining the accepted structure of a program. The substrate the other curator skills (assessment, research) attach to.
assistant-authored-study-drills
by igor1309
Orchestrate assistant-authored MCQ practice drills — provenance, cadence, difficulty shaping, coverage, two-contour balance (meaning and course nomenclature), source-grounding, and live quality repair. Defers item-quality, button delivery, and oral construction to their owning skills.
oral-exam-drills
by igor1309
Conduct oral and case-based study drills without leaking answers, over-structuring prompts, or rewarding polish over substance.
agonara-curator-research
by igor1309
Assist a human Agonara Program Curator to ground and expand a program — record external sources with justifying metadata, read coverage-gap diagnostics as draft guidance, and generate, trace, and explicitly accept or reject LLM-assisted curation suggestions. Report sources, gaps, and suggestions only from live API reads, never from memory; present suggestions as a numbered list and act only on the item numbers the curator picks after verifying origin. Use when a curator's assistant agent is researching a program: adding sources, finding gaps, or reviewing generated suggestions. Suggestions and gaps are draft material until accepted; accepting a suggestion only materializes a draft candidate, never accepted corpus.
agonara-curator-assessment
by igor1309
Assist a human Agonara Program Curator to author assessment — open-answer checks, quick checks, and readiness checks via the candidate→acceptance→revision flow, define program-owned assessment criteria, link checks to learner-state targets, and keep answer keys and correctness curator-only (never student-facing). Treat your own drafts and reports as untrusted: present a numbered list, the curator verifies each candidate (and its answer key) against trusted state and approves by explicit item number before you accept, acceptance is per item not batched, and any acceptance lacking that verification is contaminated. Use when a curator's assistant agent is adding or accepting checks or criteria on a program. Builds on the corpus skill; checks attach to accepted corpus entities.
implementation-plan
by igor1309
Use when authoring, editing, or validating implementation plans under any docs/plans path; defines task/phase/step vocabulary, required execution declarations, and the acceptance checklist that catches invalid worktree, branch, and push-policy declarations before execution starts.