1102tools Claude Skills

Claude Skills for federal acquisition professionals.

Architecture diagram showing how each instrument chains scope, pricing, and data sources. FAR contracts: SOW/PWS Builder feeds three IGCE Builders (FFP, LH/T&M, Cost-Reimbursement) pulling from BLS OEWS, GSA CALC+, and GSA Per Diem. Other Transactions: OT Project Description Builder feeds OT Cost Analysis pulling from the same three data sources.

Before you build: Not every acquisition capability should be an AI tool. Dozens of potential skills were evaluated and several were intentionally excluded because they cross the line from data assembly into professional judgment.

See AI-BOUNDARIES.md.

Skills

API Data Sources

Skill	Key	Description
USASpending API	No key	Federal contract and award data. PIIDs, vendor awards, transaction histories, agency spending. Reference content merged into main skill.
GSA CALC+ Ceiling Rates	No key	Awarded NTE hourly rates from GSA MAS contracts (230K+ records). Independently tested across 8 runs on 2 Claude models (112 of 112 assertions passed; Round 2 patches shipped and validated). Testing Record. Reference content merged into main skill.
BLS OEWS Wages	BLS key	Market wage data covering ~830 occupations across 530+ metro areas. Independently tested across 16 runs in two waves and 2 Claude models (112 of 112 assertions passed; Rounds 2 through 4 patches shipped; Dayton MSA renumbering + probe series verified live). Testing Record. Reference content merged into main skill.
GSA Per Diem Rates	api.data.gov	Federal travel per diem (lodging + M&IE) for all CONUS locations. Reference content merged into main skill.
SAM.gov API	SAM.gov	Entity registration (UEI/CAGE), exclusion/debarment records, contract opportunities, contract awards (FPDS replacement).
SAM.gov Reference	No key	Entity and award schemas, business type codes, composite workflows. Install alongside main skill.

Orchestration Skills

Skill	Key	Requires	Description
SOW/PWS Builder	No key	--	Structured scope decision tree producing contract-file-ready SOW or PWS. FAR 37.102(d) compliant: staffing handoff for the IGCE Builder is delivered as chat output, never embedded in the document body. Independently tested across 14 runs in two waves and 2 Claude models (163 of 168 assertions passed; 5 patches shipped). Testing Record.
IGCE Builder: FFP	No key*	BLS OEWS, GSA CALC+, GSA Per Diem	Firm-fixed-price IGCEs with layered wrap rate buildup (fringe, overhead, G&A, profit). Independently tested across 15 runs in three waves on 2 Claude models (42 of 42 Wave 3 retest assertions passed on Opus after Round 5; Rounds 5 and 6 patches shipped). Testing Record.
IGCE Builder: LH/T&M	No key*	BLS OEWS, GSA CALC+, GSA Per Diem	Labor Hour and T&M IGCEs with burden multiplier pricing.
IGCE Builder: Cost-Reimbursement	No key*	BLS OEWS, GSA CALC+, GSA Per Diem	CPFF, CPAF, CPIF IGCEs with fee structure analysis and statutory fee caps.

Other Transaction (OT) Skills

Skill	Key	Requires	Description
OT Project Description Builder	No key	--	Milestone-based project descriptions for prototype OT agreements under 10 USC 4021/4022. Replaces the SOW/PWS for OTs: structures work around TRL progression phases and go/no-go gates instead of task/subtask CLINs. Handles NDC, small business, traditional (with cost sharing), and consortium-brokered agreements. Produces a .docx agreement attachment and a chat-only milestone handoff table for the OT Cost Analysis.
OT Cost Analysis	No key*	BLS OEWS, GSA CALC+, GSA Per Diem	Should-cost estimates and price reasonableness analyses for OT agreements. Milestone-based pricing citing 10 USC 4021 instead of FAR 15.404. Handles cost-sharing math (10 USC 4022(d)), consortium management fees, fixed-price and cost-type milestone payments, and pre-solicitation budget planning. Produces a formula-driven .xlsx workbook with scenario analysis and a price reasonableness memo for the agreement file.

*Uses keys from other installed skills. No additional key needed.

MCP Servers

Want more consistent results across runs? Install the MCP servers. Each of the 5 API data source skills above also ships as a Model Context Protocol (MCP) server on PyPI. When installed in Claude Desktop or another MCP client, orchestration skills call these servers directly for deterministic tool calls instead of letting Claude generate API request code through SKILL.md instructions alone.

What is MCP?

Model Context Protocol is Anthropic's open standard for LLM applications to communicate with external tools and data sources. MCP servers run locally, expose a set of callable tools, and return structured responses. They work in any MCP-compatible client including Claude Desktop, Claude Code, Cursor, Cline, Continue, and Zed.

Why use the MCP servers?

Deterministic tool calls. Every call to search_awards() produces the same structured output, rather than Claude generating slightly different API request code across runs.
Orchestration skills prefer them. If the SAM.gov MCP is installed, the SOW/PWS Builder and IGCE Builder skills use it for vendor lookups instead of falling back to API-call generation through SKILL.md.
Independently hardened. Each server went through 3 to 6 audit rounds with live testing against the production API. See each TESTING.md for the full record.
Published to PyPI with Trusted Publisher. Install via uvx with no manual Python environment setup.

Install

Install Claude Desktop on Mac or Windows.
Open claude_desktop_config.json:
- Mac: ~/Library/Application Support/Claude/claude_desktop_config.json
- Windows: %APPDATA%\Claude\claude_desktop_config.json
Paste an mcpServers block for each server you want. Example config for all 5:

{
  "mcpServers": {
    "bls-oews": {
      "command": "uvx",
      "args": ["bls-oews-mcp"],
      "env": {"BLS_API_KEY": "your-key-here"}
    },
    "gsa-calc": {
      "command": "uvx",
      "args": ["gsa-calc-mcp"]
    },
    "gsa-perdiem": {
      "command": "uvx",
      "args": ["gsa-perdiem-mcp"],
      "env": {"PERDIEM_API_KEY": "your-key-here"}
    },
    "sam-gov": {
      "command": "uvx",
      "args": ["sam-gov-mcp"],
      "env": {"SAM_API_KEY": "your-key-here"}
    },
    "usaspending": {
      "command": "uvx",
      "args": ["usaspending-gov-mcp"]
    }
  }
}

Restart Claude Desktop.
Orchestration skills will use the installed MCP servers automatically when available.

See API Keys below for where to get free keys for the 3 servers that need them. gsa-calc and usaspending need no key.

All 5 MCP servers

Each repo ships with a TESTING.md documenting its hardening record. Click through for bug counts, audit rounds, and signature bug stories.

MCP	Version	Tests	Source	Testing Record
bls-oews-mcp	0.2.2	60	GitHub	Testing Record
gsa-calc-mcp	0.2.2	117	GitHub	Testing Record
gsa-perdiem-mcp	0.2.1	172	GitHub	Testing Record
sam-gov-mcp	0.3.1	79	GitHub	Testing Record
usaspending-gov-mcp	0.2.3	62	GitHub	Testing Record

Across all 5 MCPs: 490 regression tests covering roughly 224 bugs fixed during hardening.

Testing and Validation

Skills in this repo are progressively being run through independent end-to-end validation. Each testing wave uses the following methodology:

Scenarios are run on claude.ai web chat, the same environment real users run in.
Each worker output is graded by a separate Claude instance (Claude Code Opus 4.7 with 1M context, max effort) reading only the final artifact and chat transcript, with no access to the worker's drafting reasoning.
Grading is binary pass/fail against a 14-point assertion matrix covering FAR compliance, document structure, voice consistency, and handoff format.
Findings that surface real bugs or consistent behavioral gaps become skill patches.
Every tested skill ships with a public Testing Record documenting methodology, results, and changes made.

Orchestration Skills Testing Records

Skill	Scenarios	Models tested	Assertion result	Testing Record
SOW/PWS Builder	14 runs (2 waves)	Opus 4.7 + Sonnet 4.6	163 of 168 passed, 5 patches shipped	Testing Record
IGCE Builder: FFP	15 runs (3 waves)	Opus 4.7 + Sonnet 4.6	42 of 42 Wave 3 retest assertions passed on Opus after Round 5, Rounds 5 and 6 patches shipped	Testing Record
IGCE Builder: LH/T&M	Pending	--	--	Not yet published
IGCE Builder: Cost-Reimbursement	Pending	--	--	Not yet published

Other Transaction (OT) Skills Testing Records

Skill	Scenarios	Models tested	Assertion result	Testing Record
OT Project Description Builder	Pending	--	--	Not yet published
OT Cost Analysis	Pending	--	--	Not yet published

MCP Servers Testing Records

MCP servers use a different testing methodology from the skills above: instead of scenario-graded assertions, each server was hardened through multi-round live audits targeting input validation, response-shape fragility, and silent-wrong-data bugs. Each was probed against its production API with edge cases, injection attempts, boundary values, and unusual response shapes. See each TESTING.md for the round-by-round breakdown.

MCP	Audit rounds	Regression tests	Bugs fixed	Testing Record
bls-oews-mcp	5	60	22	Testing Record
gsa-calc-mcp	4	117	86	Testing Record
gsa-perdiem-mcp	6	172	55	Testing Record
sam-gov-mcp	4	79	46	Testing Record
usaspending-gov-mcp	4	62	15	Testing Record

Combined across all 5 MCPs: 490 regression tests covering roughly 224 bugs fixed.

API Keys

Three free keys cover everything:

BLS: Register at data.bls.gov/registrationEngine (500 queries/day)
api.data.gov: Register at api.data.gov/signup (1,000 req/hr; covers Per Diem)
SAM.gov: Register at SAM.gov (free account; API key in your profile; keys expire every 90 days)

Tell Claude to remember your keys and it will use them automatically.

Install

Download and unzip a skill
In Claude: Customize > Skills > + > Create skill > Upload a skill > drag in each SKILL.md file individually
For skills with a reference companion, install both
Ask a question naturally. Claude reads the instructions and makes the API call.

License

MIT