Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add lugassawan/swe-workbench --skill principle-performancegit clone --depth 1 https://github.com/lugassawan/swe-workbenchWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/lugassawan/swe-workbench/principle-performance)<a href="https://agentmods.dev/skills/lugassawan/swe-workbench/principle-performance"><img src="https://agentmods.dev/badge/skills/lugassawan/swe-workbench/principle-performance.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00092 | $0.01612 |
| Opus 5 | $0.00046 | $0.00806 |
| Sonnet 5 | $0.00018 | $0.00322 |
| Haiku 4.5 | $0.00009 | $0.00161 |
Grade A, and why
principle-performance scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 96 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Performance
Performance bugs are design bugs. They are cheapest to fix before the first line of code is written. This skill teaches design-time discipline — choosing the right algorithm, data structure, and access pattern — not runtime profiler operation.
Latency vs Throughput
They pull in opposite directions; name the goal before optimizing.
- Latency: time to serve one request. Throughput: requests served per unit time. Improving one often degrades the other.
- Tail latency (p99, p999) is a separate budget from mean latency — do not let averages hide outliers.
- Batching and buffering improve throughput at the cost of per-item latency; state this trade-off explicitly.
- Choose the objective first: a real-time API and a batch pipeline have different success criteria.
Profile Before You Optimize
Measurement beats intuition; no fix without a hot path identified by data.
- Identify the bottleneck with a profiler before changing code — optimizing a path that accounts for 5% of runtime cannot yield more than a 5% total improvement, no matter how perfect the fix.
- Benchmark before and after each change; a "feels faster" claim is not evidence.
- Most code is cold; optimize only the identified hot path. Premature optimization is applied to the wrong place.
- A profile that surprises you is information; a profile you skipped is a bug waiting to be filed.
Big-O Where It Bites
Algorithmic complexity matters when N grows; the right abstraction is cheaper than any constant-factor tweak.
- Nested loops over collections are O(n²) by default — verify that the outer N is bounded and small.
- Membership tests on lists are O(n); use a hash set when the check is inside a loop.
- String concatenation inside a loop builds O(n²) bytes; accumulate then join once outside the loop.
- Sort once, query many times — precompute sorted order or indexes when access patterns allow.
- Accidental quadratic is the most common performance regression; review any loop whose body touches a collection.
What ships with it
10 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- examples/cache-aside.cs.md 2.8 KB
- examples/cache-aside.go.md 2.9 KB
- examples/cache-aside.java.md 2.6 KB
- examples/cache-aside.kt.md 2.4 KB
- examples/cache-aside.py.md 2.3 KB
- examples/cache-aside.rb.md 1.9 KB
- examples/cache-aside.rs.md 3.3 KB
- examples/cache-aside.swift.md 2.7 KB
- examples/cache-aside.ts.md 2.2 KB
- triggers.txt 553 B
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 3d ago First seen · 96 lines · 92 tokens per session scan A 7826127784f9
principle-performance is a skill published in the GitHub repository lugassawan/swe-workbench (2 stars, last pushed yesterday), licensed MIT. It adds 92 tokens to every session and 1,612 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-03.
Other skills, from other repositories
code-forge
Generate implementation code from an approved design blueprint or verbal requirements. Composes context anchoring, architecture, clean code, DDD, security, and test quality into an inside-out implementation workflow. Use when moving from design to code, implementing approved contracts, or when the user says…
slack-block-kit
Proactively apply when generating Slack API payloads with blocks, chat.postMessage calls with structured content, streaming AI responses, or views.open/views.publish calls. Triggers on Block Kit, Slack blocks, section block, actions block, header block, context block, alert block, card block, carousel block, container…
clean-ddd-hexagonal
Proactively apply when designing APIs, microservices, or scalable backend structure. Triggers on DDD, Clean Architecture, Hexagonal, ports and adapters, entities, value objects, domain events, CQRS, event sourcing, repository pattern, use cases, onion architecture, outbox pattern, aggregate root, anti-corruption…
postgres-drizzle
Proactively apply when creating APIs, backends, or data models. Triggers on PostgreSQL, Postgres, Drizzle, drizzle-orm, drizzle-kit, database, schema, pgTable, tables, columns, indexes, queries, migrations, ORM, relations, relational queries, joins, transactions, SQL, connection pooling, PgBouncer, N+1, JSONB, RLS…
functions-development
Build serverless Go or Python functions for Falcon Foundry apps. TRIGGER when user asks to "create a function", "write a serverless function", "build backend logic", runs foundry functions create, or needs help with FDK handler patterns, function testing, or collection integration from functions. Also TRIGGER when…
go-expert
Use when writing or reviewing Go backend services - go.mod, .go files, net/http handlers, pgx/sqlc database code, goroutines, context.Context plumbing, or failing go test runs. Builds HTTP APIs on the Go 1.22+ stdlib router, fixes error-wrapping and context-cancellation bugs, designs leak-free goroutine lifecycles…