transparency-report

A reusable agent skill for adding a compact provenance / source-attribution block to answers.

This skill is designed for cases where the user asks:

• where the answer came from
• what context influenced the answer
• which files, tools, memory, web results, or prompts were used
• how the assistant arrived at the result

It appends a short transparency section with a Markdown table that shows:

• which context categories were used
• which were possibly used
• which were not used

What it does

The skill defines a structured provenance block that:

• adapts to the language of the answer
• works for Russian and English out of the box
• distinguishes between direct evidence and possible contextual influence
• keeps the output compact and scannable

Typical output includes:

• a localized heading
• a short disclaimer
• a 3-column Markdown table:
  • Category
  • Status
  • Note

Included file

• SKILL.md — the main skill definition and behavior rules

Suggested use cases

Use this skill when:

• a user asks for provenance, traceability, or transparency
• a user wants to know what influenced the answer
• you are debugging agent behavior
• you want explicit source visibility in generated responses

Do not append the provenance block to normal answers unless the user asks for it.

Status model

The skill uses exactly one status per row:

• Used
• Possibly used
• Not used

Covered categories

The skill defines a comprehensive provenance table covering:

• User message
• Current chat context
• Images / visual input
• Project files / uploaded files
• System prompt / behavior rules
• User preferences / style
• Project instructions
• Location and time
• Memory
• Past chats
• Web search
• Web fetch / URL pages
• Connected apps / MCP
• Skills
• Tool calls and results
• Code execution / sandbox
• Model knowledge
• Reasoning and assumptions

Repository layout

.
├── SKILL.md
├── README.md
└── LICENSE

License

MIT

Author

Sergei Grieg