个人作者
Milton Adina
CS/Cybersecurity SWE & former Epidemiologist. Building secure, test-driven full-stack & mobile apps (TS/React/Node, Java/Spring, Flutter).
- Skills
- 30
- 仓库
- 1
- 职业覆盖
- 0
- 场景覆盖
- 0
- Stars
- ★ 0
Milton Adina 的 Skills 30
karpathy-guidelines
Behavioral guidelines that reduce common LLM coding mistakes. Apply whenever writing, modifying, or reviewing code. Don't assume requirements, don't overcomplicate, don't refactor unrelated code, and define success criteria before implementing. Use this skill on every non-trivial coding task to avoid bloat, silent assumptions, and completion-bias.
unsigned-missing-sig-fixture
Test fixture (F.09) — SKILL.md without an accompanying .sig file. Default install rejects with exit 2; --allow-unsigned --rationale="..." proceeds.
unsigned-mismatched-sig-fixture
Test fixture (F.09) — SKILL.md whose accompanying .sig signs different content. NFR-F3 hash check catches this even with --allow-unsigned.
karpathy-guidelines
Behavioral guidelines that reduce common LLM coding mistakes. Apply whenever writing, modifying, or reviewing code. Don't assume requirements, don't overcomplicate, don't refactor unrelated code, and define success criteria before implementing. Use this skill on every non-trivial coding task to avoid bloat, silent assumptions, and completion-bias.
webhook-idempotency
Defend against double-processing of webhook events across every provider (Stripe, GitHub, Slack, Twilio, SendGrid). Idempotency-key pattern + replay-window enforcement + dedupe-storage strategy. Triggers on any webhook handler authoring (route matching /webhook(s)?/, Stripe-Signature / X-Hub-Signature-256 header references). Defense against OWASP ASI02 (Tool Misuse) -- a replayed webhook that re-triggers a destructive tool is the canonical control-flow-level replay attack. Universal across providers; stack-specific signature verification lives in stack-specific skills.
semgrep-scan
Run Semgrep static analysis to detect insecure code patterns - SQL injection, XSS, path traversal, weak crypto, broken auth, OWASP Top 10 issues. Use on every file change touching security-sensitive code and as part of the tiered security gate. Complements gitleaks (which catches secrets) by catching insecure logic.
prompt-injection-defense
Sanitize all external content before it enters the agent's context window. Use whenever the agent ingests documents, web pages, RAG results, tool outputs, file contents, emails, or any data from outside the user's direct typed input. Defense against OWASP ASI01 (Goal Hijacking) and ASI04 (Indirect Prompt Injection) - the
owasp-asi-threat-model
Produce a STRIDE + OWASP Top 10 for Agentic Applications (ASI) 2026 threat model for any new feature, API, or system. Use whenever designing a feature that touches auth, data, agents, tools, or external inputs. Triggers on terms like "design", "architecture", "new endpoint", "new agent", "new MCP", or "security review". The ASI 2026 framework (released Dec 2025) is the industry standard for agentic AI security.
gitleaks-scan
Run gitleaks against the working tree to detect committed secrets, API keys, tokens, and credentials. Use whenever a file has been written, before any git commit, before any push, and as part of the tiered security gate. Catches the most common cause of credential leaks - accidental commits.
spec-extraction
Convert a user's prose requirements into EARS-formatted specifications under /specs/. Use whenever the user describes new functionality, requirements, or a feature without an existing spec. Required first step for any non-trivial implementation - no spec, no work. Triggers on phrases like "add feature X", "build a Y", "we need to do Z", "the system should...".
session-summary
Generate the end-of-session human-reviewable report. Use at session end before writing the baton. Aggregates all verified claims, lists open blockers, links proofs, and surfaces low-confidence items for human review. Required for any session that produced verifiable work.
proof-of-work
Produce structured proof artifacts for every completed claim. Use this skill whenever the agent claims to have implemented something, tested something, deployed something, or completed a task. Every claim must ship with proof — git SHA, files changed, test command, exit code, output tail, and a reproducibility hash. Claims without proof are rejected by the claim-validator. Triggers on any output that asserts completion.
plan-decomposition
Break an approved spec into atomic, ordered, verifiable tasks before any code is written. Use after spec-extraction confirms a spec is approved and before invoking goal-loop. Each task has its own success criteria, estimated effort, and explicit dependencies. Prevents both "where do I start" paralysis and "ship-and-pray" rushes.
multi-tool-failover
Manage session continuity when switching between Claude Code, Codex CLI, Cursor, Antigravity, Kiro, or local LLMs (OpenCode, Aider). Use when approaching session limits, switching to a different tool's strengths, or when one tool fails and another must continue. Relies on the baton protocol and universal SKILL.md format.
goal-loop
Execute the spec-tests-implement-verify cycle for one atomic task at a time. Use after plan-decomposition has produced a task list. This is the workhorse loop - the agent implements, runs tests, reflects on failure, retries with bounded iteration. Halts at the iteration ceiling, and when a budget, loop-detection or scratchpad-stasis hook fires where the host wires one. See constitution/LOOP.md for full protocol.
baton-handoff
Write a complete handoff document at session end so the next agent (Claude Code, Codex, Cursor, Antigravity, Kiro, or local LLM) resumes exactly where this one stopped. Use whenever a session is approaching its limit, the user types /checkpoint, or the agent recognizes natural completion of a task. Surgical honesty - no context loss across tool boundaries.
observability-instrument
Add OpenTelemetry spans with user_id and tenant_id in baggage to any new code path that handles requests, runs agents, or performs billable operations. Use whenever creating a new API endpoint, background job, agent invocation, or external API call. Inline PII redaction at the span exporter is mandatory before any production deploy.
cost-attribution
Track per-client and per-feature LLM and cloud spend with cryptographic audit trail. Use whenever billable operations occur - LLM API calls, infrastructure provisioning, batch jobs. Required for any project with multiple clients or cost-sensitive operations. Integrates with Stratum's append-only billing_records table.