Skills in this repository
SkillsMP has collected 98 skills from alexandremendoncaalvaro/agentic-development. Open a skill to review its source and details.
alexandremendoncaalvaro/agentic-developmentShowing 40 of 98 collected skills.
Four-source research before non-trivial code — official docs, validated references, in-repo patterns, git history; synthesizes the happy path, persists an evidence record, gates deviations. Use before implementing, choosing a library or pattern, or on "ground…
Check a pull request (CI, review, mergeability), merge once the owner confirms. Use on "merge this PR".
Open a pull request in the repo body shape once the owner approves the draft. Use on "open a PR".
Two-axis fresh-context code review (Standards against AGENTS.md and ADRs; Spec against the originating task) run on Codex as one inline pass and audit trail, findings side by side, no approve verdict. Use before merging a diff, branch, or PR; "review this",…
Four-source research before non-trivial code — official docs, validated references, in-repo patterns, git history; synthesizes the happy path, persists an evidence record, gates deviations. Use before implementing, choosing a library or pattern, or on "ground…
Check a pull request (CI, review, mergeability), merge once the owner confirms. Use on "merge this PR".
Open a pull request in the repo body shape once the owner approves the draft. Use on "open a PR".
Maximum-gate adversarial audit of a diff, branch, PR, or drafted claims against the repo's rule-set — on Codex one inline pass and audit trail, every rule given a verdict. Use before posting to the team, opening a PR, or handing off; "audit against the…
Two-axis fresh-context code review (Standards against AGENTS.md and ADRs; Spec against the originating task) run on Codex as one inline pass and audit trail, findings side by side, no approve verdict. Use before merging a diff, branch, or PR; "review this",…
Two-axis fresh-context code review (Standards against AGENTS.md and ADRs; Spec against the originating task) run as parallel subagents, findings side by side, never an approve verdict. Use before merging a diff, branch, or PR; "review this", "review…
Two-axis fresh-context code review (Standards against AGENTS.md and ADRs; Spec against the originating task) run as parallel subagents, findings side by side, never an approve verdict. Use before merging a diff, branch, or PR; "review this", "review…
Design, interpret, refine, and audit decision-focused evaluations. Use when asking how to evaluate a change, define measures or decision rules, test fit for purpose, interpret results, or audit evaluation claims. Routes open research, uncertain techniques,…
Design, interpret, refine, and audit decision-focused evaluations. Use when asking how to evaluate a change, define measures or decision rules, test fit for purpose, interpret results, or audit evaluation claims. Routes open research, uncertain techniques,…
Create a read-only decision-maker brief from live project and session state or a settled fact packet. Use when the user says "brief me", "give me a summary", "catch me up", "where are we?", asks what is happening now and why, or needs a plain-language boss…
Survey the project's artifact stack and recommend prioritized next actions, like flutter doctor. Use when unsure where the work stands; "what's next", "where am I", "project status", "doctor", "what should I do". Read-only.
Resolve repository and configured GitHub evidence into a bounded, provenance-bearing project-state packet. Use when ad-next, project ad-roadmap, or standalone ad-brief needs current work from local artifacts plus issues or pull requests; "project sources",…
Read-only roadmap report — reconciles roadmap tiers against specs and tasks, then explains the main delivery front, current work, next front, blockers, and progress as a plain-language Markdown checklist. Use on "how much is done", "are we on track",…
Create a read-only decision-maker brief from live project and session state or a settled fact packet. Use when the user says "brief me", "give me a summary", "catch me up", "where are we?", asks what is happening now and why, or needs a plain-language boss…
Survey the project's artifact stack and recommend prioritized next actions, like flutter doctor. Use when unsure where the work stands; "what's next", "where am I", "project status", "doctor", "what should I do". Read-only.
Resolve repository and configured GitHub evidence into a bounded, provenance-bearing project-state packet. Use when ad-next, project ad-roadmap, or standalone ad-brief needs current work from local artifacts plus issues or pull requests; "project sources",…
Read-only roadmap report — reconciles roadmap tiers against specs and tasks, then explains the main delivery front, current work, next front, blockers, and progress as a plain-language Markdown checklist. Use on "how much is done", "are we on track",…
Read-only drift audit of AGENTS.md, ARCHITECTURE.md, ADR statuses, specs, and documentation discipline against what the code does. Use on "audit the docs", "check for drift", "are the docs stale", "sanity-check the ADRs". Reports a drift list; writes nothing.
Read-only drift audit of AGENTS.md, ARCHITECTURE.md, ADR statuses, specs, and documentation discipline against what the code does. Use on "audit the docs", "check for drift", "are the docs stale", "sanity-check the ADRs". Reports a drift list; writes nothing.
Resolve the host's global and project rules for inspection or correction. Use when the user asks which rules apply, says "The Rules", asks to refresh the rules, or corrects workflow discipline. Inventory mode lists topics; correction mode reapplies…
Resolve the host's global and project rules for inspection or correction. Use when the user asks which rules apply, says "The Rules", asks to refresh the rules, or corrects workflow discipline. Inventory mode lists topics; correction mode reapplies…
Scaffold a staged spike with golden fixtures when the spec is clear but the technique is uncertain — discovery, fixture, gated pipeline with per-stage debug artifacts, two-layer evaluation. Use on "spike", "which library", "evaluate approaches", "golden…
Maximum-gate adversarial audit of a diff, branch, PR, or drafted claims against the repo's rule-set — one fresh reviewer per rule-group, every rule given a verdict. Use before posting to the team, opening a PR, or handing off; "audit against the rules",…
Find deepening opportunities (Ousterhout/Feathers — module, interface, depth, seam, adapter) and propose refactors with deletion-test framing. Use when code feels shallow or pass-through; "deepen", "refactor for depth", "shallow module", "deletion test",…
De-risk a defined task before building — list unknowns in a risk register, retire each with ad-ground, ad-spike, ad-tdg, or ad-grill-me, stop when technical risk falls below non-technical risk. Use on "de-risk", "before I build", "is this safe to build",…
Disciplined diagnosis for hard bugs and performance regressions — build a feedback loop, reproduce, rank 3-5 falsifiable hypotheses, instrument, fix with a regression test. Use on "this is broken", "debug this", "find the bug", "performance regression",…
Outcome-based implementation — given input and expected output, generate three strategies and pick one by a single criterion; Test Dependency Map before modifying. Use when the outcome is known but the strategy is not; "TDG", "three approaches", "pick by…
Maximum-gate adversarial audit of a diff, branch, PR, or drafted claims against the repo's rule-set — one fresh reviewer per rule-group, every rule given a verdict. Use before posting to the team, opening a PR, or handing off; "audit against the rules",…
Turn an open question into an evidence-graded study at doc/research/NNNN-<slug>.md with a defensible conclusion, not code. Use on "should we adopt X", "A vs B", "state of the art", "evaluate options", "literature review", "research this". Graduates binding…
Turn an open question into an evidence-graded study at doc/research/NNNN-<slug>.md with a defensible conclusion, not code. Use on "should we adopt X", "A vs B", "state of the art", "evaluate options", "literature review", "research this". Graduates binding…
Compact the session into a handoff a fresh agent can resume from, saved to the OS temp dir with live state, open decisions, and next skills. Use when context runs low or on "handoff", "wrap up the session", "save context for next session", "pass to another…
Compact the session into a handoff a fresh agent can resume from, saved to the OS temp dir with live state, open decisions, and next skills. Use when context runs low or on "handoff", "wrap up the session", "save context for next session", "pass to another…
Interview before research — challenge a fuzzy ask against the codebase, one question at a time with a recommended answer, until the decision tree is resolved. Use when the ask is vague, bundles decisions, or lacks a success condition; "grill me", "challenge…
Universal posture for non-trivial work — think before coding, decide when grounded, ground in real patterns, simplicity, surgical changes, verifiable goals, verify before claiming done, report for a decision-maker, documentation discipline. Auto-loads on…
Draft a feature specification at doc/specs/NNNN-<slug>.md (Spec Kit-aligned — who, what, measurable success criteria, non-goals), inheriting scope from the PRD. Use when a feature needs a contract before tasks; "write a spec", "feature spec", "acceptance…
Draft a task file at doc/tasks/NNNN-<slug>.md with checkbox acceptance criteria and an append-only Notes log. Use when work needs a tracked unit; "create a task", "open a ticket", "work item", "backlog entry", "plan this as a task". Status starts proposed.