This v1.6.0 bundle includes small local Python scripts in addition to Markdown and JSON templates.

• Markdown skill instructions
• JSON and Markdown templates
• small local Python scripts
• examples and reference notes
• release metadata files

• embedded credentials
• third-party Python dependencies
• automatic email sending
• automatic proposal submission

The scripts in scripts/ should:

• use the Python standard library only
• avoid dynamic code execution
• avoid shelling out except where the operator explicitly allows it
• write only inside the active workspace

• never print SAM_API_KEY
• never persist secret values into procurement/
• never write secret fragments into drafts, evidence, or run logs

Use:

• SAM.gov
• USAspending.gov
• SBA SUBNet
• Acquisition.gov

Treat commercial sources as optional enrichment only when explicitly enabled.

Browser retrieval may be necessary, but it is less deterministic than direct API use.

Use it only when:

• official API retrieval is unavailable or incomplete
• the page is JS-wrapped
• the user explicitly asked for browser execution

Drafting emails is allowed.

Sending emails, outreach, submissions, or forms should require:

• explicit user approval
• sender identity details
• a clear send action outside the passive research path

The model should not present:

• a seed stub as a final brief
• a legacy brief as if it were fresh
• a menu-only response as completed research

• use a dedicated agent and workspace
• keep network access constrained where practical
• review script changes like code, not like prompt text
• keep a golden example set for regressions