Update to Attio
Snapshot Nov 17, 2026 · 00:10 UTC · version 1.0.15
Collection source: downloaded plugin package.
Payment or plan references changed
Skill instructions — Now says: call could match, list the candidates with title and date, and ask which one. Metadata search returns no titles, so get them from `search-meetings` with the...
Observed in published text. Live prices and checkout terms have not been verified by this change.
Up to 3 plugins free · no card required.
All fields and original values.
Skill instructions
Nov 11, 2026 · 12:08 UTC
recording could match, list the candidates with title, date and participants, and ask which one. Metadata search returns no titles, so get them from `search-meetings` over the same time range. Analyzing the wrong call wastes more of the ...
Nov 17, 2026 · 00:10 UTC
call could match, list the candidates with title and date, and ask which one. Metadata search returns no titles, so get them from `search-meetings` with the same filters and time range. Analyzing the wrong call wastes more of the user's ...
Supporting files
Nov 11, 2026 · 12:08 UTC
[{"relative_path":"references/demo.md","size_in_bytes":707},{"relative_path":"references/discovery.md","size_in_bytes":850},{"relative_path":"references/evaluation.md","size_in_bytes":670},{"relative_path":"references/negotiation.md","si...
Nov 17, 2026 · 00:10 UTC
[{"relative_path":"agents/openai.yaml","size_in_bytes":213},{"relative_path":"assets/icon-small.svg","size_in_bytes":2134},{"relative_path":"assets/icon.png","size_in_bytes":8865},{"relative_path":"references/demo.md","size_in_bytes":707...
Compare saved observations
Download comparison JSONSKILL.md text changes
− removed · + added
--- before +++ after @@ -20,16 +20,16 @@ Add a time range whenever the user gives one. Results are newest first. -If more than one recording could match, list the candidates with title, date and participants, and ask which one. Metadata search returns no titles, so get them from `search-meetings` over the same time range. Analyzing the wrong call wastes more of the user's time than one question does. +If more than one call could match, list the candidates with title and date, and ask which one. Metadata search returns no titles, so get them from `search-meetings` with the same filters and time range. Analyzing the wrong call wastes more of the user's time than one question does. If the meeting has no recording, say so. Calls logged from a dialer are activity records, not call recordings, and this skill does not cover them. ## 2. Read the full transcript -Fetch the recording with `get-call-recording` and read all of it before you form a view. The call type often becomes clear late, for example when pricing comes up near the end. +Fetch the recording with `get-call-recording` and read all of it before you form a view. If the meeting has more than one recording, read them all as one call, and say which part each quote comes from. The call type often becomes clear late, for example when pricing comes up near the end. -- If `status` is not complete, tell the user the transcript is partial. -- Work out which speakers are on the user's side (`whoami`, `list-workspace-members`) and which are external. +- Analyze only recordings with `status: finished`. If it's `preparing` or `in-progress`, tell the user the recording isn't finished yet and to try again later. If it's `failed`, say the recording failed. Then stop. +- Work out which speakers are on the user's side (`whoami` and `list-workspace-members`, unless you already have them) and which are external. `search-meetings` lists who was invited, but only the transcript shows who spoke. Use the transcript's speakers, and the participants for anyone invited who didn't speak. - Transcription garbles names of people, companies and products. Infer the intended word from context, and do not quote the error as fact. ## 3. Classify the call
Full technical diff · 2 changed fields
changed /included_files
[
{
"relative_path": "references/demo.md",
"size_in_bytes": 707
},
{
"relative_path": "references/discovery.md",
"size_in_bytes": 850
},
{
"relative_path": "references/evaluation.md",
"size_in_bytes": 670
},
{
"relative_path": "references/negotiation.md",
"size_in_bytes": 500
},
{
"relative_path": "references/other.md",
"size_in_bytes": 376
}
][
{
"relative_path": "agents/openai.yaml",
"size_in_bytes": 213
},
{
"relative_path": "assets/icon-small.svg",
"size_in_bytes": 2134
},
{
"relative_path": "assets/icon.png",
"size_in_bytes": 8865
},
{
"relative_path": "references/demo.md",
"size_in_bytes": 707
},
{
"relative_path": "references/discovery.md",
"size_in_bytes": 850
},
{
"relative_path": "references/evaluation.md",
"size_in_bytes": 670
},
{
"relative_path": "references/negotiation.md",
"size_in_bytes": 500
},
{
"relative_path": "references/other.md",
"size_in_bytes": 376
}
]changed /skill_md_contents
"---\nname: analyze-call-recording\ndescription: Analyze a recorded call in Attio. Finds the call recording, reads the full transcript, classifies the call (discovery, demo, evaluation check-in, negotiation or other) and writes an analysis suited to that type. Use whenever the user asks to analyze, review, debrief or summarize a call or meeting recording, or wants feedback on how a call went, even if they only name the meeting, the person or the company they spoke to.\n---\n\n# Analyze call recording\n\n## 1. Find the recording\n\nIn Attio, a meeting is a calendar event and a call recording is a transcript attached to a meeting. A meeting can have no recording, one recording or several (several is an edge case, e.g. when the call recorder leaves and rejoins). `get-call-recording` needs both `meeting_id` and `call_recording_id`.\n\n| The user gives | Do this |\n| ----------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------- |\n| An Attio link or both IDs | Use the two UUIDs directly |\n| A meeting title | `search-call-recordings-by-metadata` with `meeting_title_query` |\n| A person or company | `search-records` for its record ID, then `search-call-recordings-by-metadata` with `speaker_person_record_ids` (person) or `related_record_ids` (company, deal) |\n| A topic (\"the call where they raised security\") | `semantic-search-call-recordings` |\n| \"My last call\" | `whoami` (skip if you already know who the user is), then `search-call-recordings-by-metadata` with `speaker_workspace_member_ids` |\n| A date | `search-meetings`, then use its `call_recording_ids` |\n\nAdd a time range whenever the user gives one. Results are newest first.\n\nIf more than one recording could match, list the candidates with title, date and participants, and ask which one. Metadata search returns no titles, so get them from `search-meetings` over the same time range. Analyzing the wrong call wastes more of the user's time than one question does.\n\nIf the meeting has no recording, say so. Calls logged from a dialer are activity records, not call recordings, and this skill does not cover them.\n\n## 2. Read the full transcript\n\nFetch the recording with `get-call-recording` and read all of it before you form a view. The call type often becomes clear late, for example when pricing comes up near the end.\n\n- If `status` is not complete, tell the user the transcript is partial.\n- Work out which speakers are on the user's side (`whoami`, `list-workspace-members`) and which are external.\n- Transcription garbles names of people, companies and products. Infer the intended word from context, and do not quote the error as fact.\n\n## 3. Classify the call\n\nPick one primary type and read its reference file.\n\n| Type | Signals | Reference |\n| ------------------- | ------------------------------------------------------------------------------------------------------------ | --------------------------- |\n| Discovery | The seller asks about the buyer's situation, problems and process. Little or no product is shown. | `references/discovery.md` |\n| Demo | The seller walks through the product and the buyer reacts. | `references/demo.md` |\n| Evaluation check-in | The buyer is already trialing, piloting or reviewing. The call covers progress, blockers and open questions. | `references/evaluation.md` |\n| Negotiation | Pricing, terms, contract, procurement or the final decision. | `references/negotiation.md` |\n| Other | Anything that is not a sales conversation with a prospect, or a call too unclear to place. | `references/other.md` |\n\nCalls often mix types, such as discovery that turns into a demo. Choose the type that took most of the call and mention the secondary one. State the type and your reason first, so the user can correct you before reading on.\n\n## 4. Write the analysis\n\n```\n**[Meeting title]** · [date] · [participants, grouped by side]\n**Call type:** [type] ([one-line reason])\n\n## Summary\n[2 to 4 sentences: what the call was for and where it landed]\n\n[Sections from the reference file]\n\n## Next steps\n- [Owner]: [action] ([date, if one was stated])\n\n## Gaps and risks\n- [What should have come up and did not, or signals that put the outcome at risk]\n```\n\n- **Ground every point in the transcript.** For the points that matter most, quote a short phrase with its timestamp, like \"I have a budget\" (8:54).\n- **Keep it short.** The user should be able to act on it in a minute. Leave out a section that has nothing in it, and put the missing topic under Gaps and risks.\n- **Use the user's methodology.** Do not impose a sales framework. If the user names one, such as MEDDICC or BANT, structure the analysis around it.\n\n## 5. Offer to save\n\nDo not write to Attio unless the user asks. At the end, offer to save the analysis as a note on the related record (`create-note`) and the next steps as tasks (`create-task`).\n"
"---\nname: analyze-call-recording\ndescription: Analyze a recorded call in Attio. Finds the call recording, reads the full transcript, classifies the call (discovery, demo, evaluation check-in, negotiation or other) and writes an analysis suited to that type. Use whenever the user asks to analyze, review, debrief or summarize a call or meeting recording, or wants feedback on how a call went, even if they only name the meeting, the person or the company they spoke to.\n---\n\n# Analyze call recording\n\n## 1. Find the recording\n\nIn Attio, a meeting is a calendar event and a call recording is a transcript attached to a meeting. A meeting can have no recording, one recording or several (several is an edge case, e.g. when the call recorder leaves and rejoins). `get-call-recording` needs both `meeting_id` and `call_recording_id`.\n\n| The user gives | Do this |\n| ----------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------- |\n| An Attio link or both IDs | Use the two UUIDs directly |\n| A meeting title | `search-call-recordings-by-metadata` with `meeting_title_query` |\n| A person or company | `search-records` for its record ID, then `search-call-recordings-by-metadata` with `speaker_person_record_ids` (person) or `related_record_ids` (company, deal) |\n| A topic (\"the call where they raised security\") | `semantic-search-call-recordings` |\n| \"My last call\" | `whoami` (skip if you already know who the user is), then `search-call-recordings-by-metadata` with `speaker_workspace_member_ids` |\n| A date | `search-meetings`, then use its `call_recording_ids` |\n\nAdd a time range whenever the user gives one. Results are newest first.\n\nIf more than one call could match, list the candidates with title and date, and ask which one. Metadata search returns no titles, so get them from `search-meetings` with the same filters and time range. Analyzing the wrong call wastes more of the user's time than one question does.\n\nIf the meeting has no recording, say so. Calls logged from a dialer are activity records, not call recordings, and this skill does not cover them.\n\n## 2. Read the full transcript\n\nFetch the recording with `get-call-recording` and read all of it before you form a view. If the meeting has more than one recording, read them all as one call, and say which part each quote comes from. The call type often becomes clear late, for example when pricing comes up near the end.\n\n- Analyze only recordings with `status: finished`. If it's `preparing` or `in-progress`, tell the user the recording isn't finished yet and to try again later. If it's `failed`, say the recording failed. Then stop.\n- Work out which speakers are on the user's side (`whoami` and `list-workspace-members`, unless you already have them) and which are external. `search-meetings` lists who was invited, but only the transcript shows who spoke. Use the transcript's speakers, and the participants for anyone invited who didn't speak.\n- Transcription garbles names of people, companies and products. Infer the intended word from context, and do not quote the error as fact.\n\n## 3. Classify the call\n\nPick one primary type and read its reference file.\n\n| Type | Signals | Reference |\n| ------------------- | ------------------------------------------------------------------------------------------------------------ | --------------------------- |\n| Discovery | The seller asks about the buyer's situation, problems and process. Little or no product is shown. | `references/discovery.md` |\n| Demo | The seller walks through the product and the buyer reacts. | `references/demo.md` |\n| Evaluation check-in | The buyer is already trialing, piloting or reviewing. The call covers progress, blockers and open questions. | `references/evaluation.md` |\n| Negotiation | Pricing, terms, contract, procurement or the final decision. | `references/negotiation.md` |\n| Other | Anything that is not a sales conversation with a prospect, or a call too unclear to place. | `references/other.md` |\n\nCalls often mix types, such as discovery that turns into a demo. Choose the type that took most of the call and mention the secondary one. State the type and your reason first, so the user can correct you before reading on.\n\n## 4. Write the analysis\n\n```\n**[Meeting title]** · [date] · [participants, grouped by side]\n**Call type:** [type] ([one-line reason])\n\n## Summary\n[2 to 4 sentences: what the call was for and where it landed]\n\n[Sections from the reference file]\n\n## Next steps\n- [Owner]: [action] ([date, if one was stated])\n\n## Gaps and risks\n- [What should have come up and did not, or signals that put the outcome at risk]\n```\n\n- **Ground every point in the transcript.** For the points that matter most, quote a short phrase with its timestamp, like \"I have a budget\" (8:54).\n- **Keep it short.** The user should be able to act on it in a minute. Leave out a section that has nothing in it, and put the missing topic under Gaps and risks.\n- **Use the user's methodology.** Do not impose a sales framework. If the user names one, such as MEDDICC or BANT, structure the analysis around it.\n\n## 5. Offer to save\n\nDo not write to Attio unless the user asks. At the end, offer to save the analysis as a note on the related record (`create-note`) and the next steps as tasks (`create-task`).\n"
Full snapshot data
{
"description": "Analyze a recorded call in Attio. Finds the call recording, reads the full transcript, classifies the call (discovery, demo, evaluation check-in, negotiation or other) and writes an analysis suited to that type. Use whenever the user asks to analyze, review, debrief or summarize a call or meeting recording, or wants feedback on how a call went, even if they only name the meeting, the person or the company they spoke to.",
"included_files": [
{
"relative_path": "agents/openai.yaml",
"size_in_bytes": 213
},
{
"relative_path": "assets/icon-small.svg",
"size_in_bytes": 2134
},
{
"relative_path": "assets/icon.png",
"size_in_bytes": 8865
},
{
"relative_path": "references/demo.md",
"size_in_bytes": 707
},
{
"relative_path": "references/discovery.md",
"size_in_bytes": 850
},
{
"relative_path": "references/evaluation.md",
"size_in_bytes": 670
},
{
"relative_path": "references/negotiation.md",
"size_in_bytes": 500
},
{
"relative_path": "references/other.md",
"size_in_bytes": 376
}
],
"name": "analyze-call-recording",
"skill_md_contents": "---\nname: analyze-call-recording\ndescription: Analyze a recorded call in Attio. Finds the call recording, reads the full transcript, classifies the call (discovery, demo, evaluation check-in, negotiation or other) and writes an analysis suited to that type. Use whenever the user asks to analyze, review, debrief or summarize a call or meeting recording, or wants feedback on how a call went, even if they only name the meeting, the person or the company they spoke to.\n---\n\n# Analyze call recording\n\n## 1. Find the recording\n\nIn Attio, a meeting is a calendar event and a call recording is a transcript attached to a meeting. A meeting can have no recording, one recording or several (several is an edge case, e.g. when the call recorder leaves and rejoins). `get-call-recording` needs both `meeting_id` and `call_recording_id`.\n\n| The user gives | Do this |\n| ----------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------- |\n| An Attio link or both IDs | Use the two UUIDs directly |\n| A meeting title | `search-call-recordings-by-metadata` with `meeting_title_query` |\n| A person or company | `search-records` for its record ID, then `search-call-recordings-by-metadata` with `speaker_person_record_ids` (person) or `related_record_ids` (company, deal) |\n| A topic (\"the call where they raised security\") | `semantic-search-call-recordings` |\n| \"My last call\" | `whoami` (skip if you already know who the user is), then `search-call-recordings-by-metadata` with `speaker_workspace_member_ids` |\n| A date | `search-meetings`, then use its `call_recording_ids` |\n\nAdd a time range whenever the user gives one. Results are newest first.\n\nIf more than one call could match, list the candidates with title and date, and ask which one. Metadata search returns no titles, so get them from `search-meetings` with the same filters and time range. Analyzing the wrong call wastes more of the user's time than one question does.\n\nIf the meeting has no recording, say so. Calls logged from a dialer are activity records, not call recordings, and this skill does not cover them.\n\n## 2. Read the full transcript\n\nFetch the recording with `get-call-recording` and read all of it before you form a view. If the meeting has more than one recording, read them all as one call, and say which part each quote comes from. The call type often becomes clear late, for example when pricing comes up near the end.\n\n- Analyze only recordings with `status: finished`. If it's `preparing` or `in-progress`, tell the user the recording isn't finished yet and to try again later. If it's `failed`, say the recording failed. Then stop.\n- Work out which speakers are on the user's side (`whoami` and `list-workspace-members`, unless you already have them) and which are external. `search-meetings` lists who was invited, but only the transcript shows who spoke. Use the transcript's speakers, and the participants for anyone invited who didn't speak.\n- Transcription garbles names of people, companies and products. Infer the intended word from context, and do not quote the error as fact.\n\n## 3. Classify the call\n\nPick one primary type and read its reference file.\n\n| Type | Signals | Reference |\n| ------------------- | ------------------------------------------------------------------------------------------------------------ | --------------------------- |\n| Discovery | The seller asks about the buyer's situation, problems and process. Little or no product is shown. | `references/discovery.md` |\n| Demo | The seller walks through the product and the buyer reacts. | `references/demo.md` |\n| Evaluation check-in | The buyer is already trialing, piloting or reviewing. The call covers progress, blockers and open questions. | `references/evaluation.md` |\n| Negotiation | Pricing, terms, contract, procurement or the final decision. | `references/negotiation.md` |\n| Other | Anything that is not a sales conversation with a prospect, or a call too unclear to place. | `references/other.md` |\n\nCalls often mix types, such as discovery that turns into a demo. Choose the type that took most of the call and mention the secondary one. State the type and your reason first, so the user can correct you before reading on.\n\n## 4. Write the analysis\n\n```\n**[Meeting title]** · [date] · [participants, grouped by side]\n**Call type:** [type] ([one-line reason])\n\n## Summary\n[2 to 4 sentences: what the call was for and where it landed]\n\n[Sections from the reference file]\n\n## Next steps\n- [Owner]: [action] ([date, if one was stated])\n\n## Gaps and risks\n- [What should have come up and did not, or signals that put the outcome at risk]\n```\n\n- **Ground every point in the transcript.** For the points that matter most, quote a short phrase with its timestamp, like \"I have a budget\" (8:54).\n- **Keep it short.** The user should be able to act on it in a minute. Leave out a section that has nothing in it, and put the missing topic under Gaps and risks.\n- **Use the user's methodology.** Do not impose a sales framework. If the user names one, such as MEDDICC or BANT, structure the analysis around it.\n\n## 5. Offer to save\n\nDo not write to Attio unless the user asks. At the end, offer to save the analysis as a note on the related record (`create-note`) and the next steps as tasks (`create-task`).\n"
}SHA-256 of public snapshot: 410590de7d2e0759d91fb11eb1f6f024fd5a6e321540cac62b736a97dfee8a3b