SkillsBench is a benchmark for measuring how effectively AI agents use modular skills—folders containing instructions, scripts, and resources—to complete specialized tasks. It helps researchers and developers evaluate both skill quality and agent behavior, including tasks that require combining multiple skills. The catalogue’s skills and instructions are evaluated as part of this workflow.
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 benchflow-ai/skillsbench --skill comp3-packed-decimalgit clone --depth 1 https://github.com/benchflow-ai/skillsbenchWrote 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/benchflow-ai/skillsbench/comp3-packed-decimal)<a href="https://agentmods.dev/skills/benchflow-ai/skillsbench/comp3-packed-decimal"><img src="https://agentmods.dev/badge/skills/benchflow-ai/skillsbench/comp3-packed-decimal.svg" alt="Measured on agentmods" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00072 | $0.01158 |
| Opus 5 | $0.00036 | $0.00579 |
| Sonnet 5 | $0.00014 | $0.00232 |
| Haiku 4.5 | $0.00007 | $0.00116 |
Grade A, and why
comp3-packed-decimal 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 8d 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 — 110 lines — stays where its author put it; the contents beside it link to each section on GitHub.
COMP-3 / packed decimal storage
USAGE COMP-3 (a.k.a. PACKED-DECIMAL) stores a signed numeric field as
binary-coded-decimal nibbles: each byte holds two decimal digits, except the
last byte whose low nibble is the sign:
PIC S9(11)V99 USAGE COMP-3
+--+--+--+--+--+--+--+
|D1|D2|D3|D4|D5|D6|D7| 7 bytes total for 13 digits
+--+--+--+--+--+--+--+
^^-- sign nibble in the LOW half of the last byte
Bytes-on-disk = ceil((digit_count + 1) / 2). For PIC S9(11)V99 that is
ceil(14 / 2) = 7 bytes. The high nibble of the first byte holds the first
decimal digit; the implied decimal point is not stored.
Sign nibbles:
| Nibble | Meaning |
|---|---|
C |
positive (some shops emit F for unsigned, but C is standard for SIGNED) |
D |
negative |
F |
unsigned / non-S PIC |
Declaring COMP-3 in an FD
The record layout must match the byte budget. If the data in the file is
packed but you declare the field as PIC X(n) or as an unpacked decimal,
the COBOL runtime will not decode it and the value will look like binary
garbage when displayed.
FD MAST-FILE
RECORD CONTAINS 80 CHARACTERS.
01 MAST-REC.
05 M-ACCT PIC X(6).
05 M-TYPE PIC X.
05 M-STATUS PIC X.
05 M-CCY PIC X(3).
05 M-CARRY PIC S9(11)V99 USAGE COMP-3. *> 7 bytes
05 FILLER PIC X(62). *> remainder of 80
Use a runtime check: cobc -x -free -o prog PROG.cbl && ./prog and DISPLAY
one record's COMP-3 field before relying on it. Common mistakes:
- Forgetting the
USAGE COMP-3clause -> the field is read as 7 zoned ASCII characters; values look like control characters. - Wrong digit count -> the byte budget changes and subsequent fields slide.
- Moving COMP-3 directly into a
PIC X(7)workspace -> the bytes copy but no arithmetic translation happens; the value is unusable.
Carrying the value through working storage
A common defect when there is an opening balance / carry-forward column:
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.
- 8d ago First seen · 110 lines · 72 tokens per session scan A 729203ed0047
comp3-packed-decimal is a skill published in the GitHub repository benchflow-ai/skillsbench (1,748 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 72 tokens to every session and 1,158 once invoked, about $0.0004 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 2026-08-30.
Other skills, from other repositories
hotpath_init
Configure hotpath profiling in a Rust project. Adds the hotpath dependency with feature-gated setup, instruments main with hotpath::main, functions with measure/measureall, and wraps channels, mutexes, rwlocks, streams, futures, reqwest clients, axum routers and byte-level I/O with hotpath macros. Use when the user…
scala-circe-json
Working with Circe JSON library in Scala for type-safe JSON processing.
mandarin-rhyming-guide
Master modern Mandarin pinyin finals and tone systems to create accurate rhymes for classical poetry following contemporary pronunciation standards.
scala-adt-enums
Translating Python Enums and Protocols to Scala sealed traits and algebraic data types.
scala-collections-fp
Scala collections, iterators, and functional programming patterns for Python developers.
scala-option-either
Translating Python's None/Optional to Scala's Option, Either, and Try types.