channel-catchup
Skill159 starsChanged 6 months agoMITView on GitHub
Summarize recent activity across Discord, ingested email threads, WhatsApp, and TUI channels.
We split this skill into the rules it gives an agent
36 rules. About 0 are common, 29 are distinctive, and 7 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈29
Distinctive · estimate
This author's own choices. Worth a look.
7
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 30
Do not reflexively ask for scope if you can already do a useful catch-up from available context and tools.
- Distinctive: line 32
Default to the broadest safe scope you can actually resolve:
- Distinctive: line 34
If channels are explicit, use them.
- Distinctive: line 35
If the platform/thread is explicit and no count is given, use sensible defaults.
- Distinctive: line 36
If the request is broad but concrete targets are already visible in context or tool output, read them and summarize.
- Distinctive: line 37
Only ask a clarification when no concrete readable target can be resolved without guessing.
- Distinctive: line 39
When a reasonable assumption is needed, make it, do the catch-up, and state the assumption after the summary instead of blocking first.
- Distinctive: line 41
Default limits when the user did not specify them:
- Distinctive: line 45
If no timeframe is provided, summarize the most recent activity visible in those reads
- Distinctive: line 53
if the platform is clear, proceed on that platform
- Distinctive: line 54
if the target set is broad, use all concrete readable targets you can already resolve
- Distinctive: line 55
if no timeframe is given, prefer the latest visible activity
- Distinctive: line 56
if no count is given, use the default limits above
- Distinctive: line 58
Only ask a clarification when you cannot identify any concrete readable target without guessing.
- Distinctive: line 59
If you had to infer scope, note the assumption after the answer in one short line and keep moving.
- Very specific: line 65
Use message with action="read" for each target channel before summarizing.
- Very specific: line 66
Prefer explicit channel IDs.
- Distinctive: line 67
Read each channel separately, then merge findings into one summary.
- Distinctive: line 68
Do not say you need pasted messages if you can read the requested Discord channels directly.
- Distinctive: line 69
If the user names a Discord server but not channels, default to all concrete readable channels you can already resolve from context, tool …
- Distinctive: line 70
If the ask is broad and you can resolve multiple channels, prefer covering more channels over asking for narrower scope.
- Distinctive: line 71
If channels are explicit but timeframe is missing, read the recent bounded sample and summarize it as the latest activity instead of …
- Distinctive: line 72
Ask only if you still cannot identify any concrete readable channels without guessing.
- Very specific: line 86
Use message with action="read" for the current ingested email thread or an explicit email address.
- Very specific: line 87
If already in email context, omit channelId to read the current thread.
- Distinctive: line 89
Do not claim mailbox-wide unread inbox access.
- Distinctive: line 104
If the current session context already contains the relevant conversation, summarize that directly.
- Very specific: line 106
Only use session_search if the user wants topic-based recall and accepts approximate recall rather than a strict "last N messages" summary.
- Distinctive: line 111
If the needed content is already in the current session context, summarize it directly.
- Distinctive: line 113
Do not pretend you can fetch arbitrary historical TUI messages when no direct read path exists.
- Distinctive: line 117
After reading the scoped content:
- Distinctive: line 123
do not quote large message dumps unless the user asks
- Distinctive: line 132
If the user asks for "short", keep it to a tight executive summary.
- Distinctive: line 136
Never summarize unseen content as if you read it.
- Very specific: line 137
Do not use session_search as a substitute for chronological channel reads when the user explicitly asked for the last N messages from …
- Very specific: line 141
If replying by email, append a polished corporate signature block derived from IDENTITY.md details already loaded in context.
≈ “Common” is an estimate: at least 1 in 5 of the 308,232 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Channel Catchup
Creating a concise catch-up or summary of channel content.
Examples:
- "Summarize the last 50 messages from
#announcementsand#engineering." - "Give me the important posts from our HybridAI Discord server today."
- "Catch me up on this email thread."
- "Summarize what happened in the current TUI session."
Default To Action
Do not reflexively ask for scope if you can already do a useful catch-up from available context and tools.
Default to the broadest safe scope you can actually resolve:
- If channels are explicit, use them.
- If the platform/thread is explicit and no count is given, use sensible defaults.
- If the request is broad but concrete targets are already visible in context or tool output, read them and summarize.
- Only ask a clarification when no concrete readable target can be resolved without guessing.
When a reasonable assumption is needed, make it, do the catch-up, and state the assumption after the summary instead of blocking first.
Default limits when the user did not specify them:
- Discord: last 50 messages per resolved channel
- Email: last 20 messages from the current or explicit ingested thread
- If no timeframe is provided, summarize the most recent activity visible in those reads
Scope Resolution
Resolve scope aggressively instead of asking for it.
Use these defaults:
- if the platform is clear, proceed on that platform
- if the target set is broad, use all concrete readable targets you can already resolve
- if no timeframe is given, prefer the latest visible activity
- if no count is given, use the default limits above
Only ask a clarification when you cannot identify any concrete readable target without guessing. If you had to infer scope, note the assumption after the answer in one short line and keep moving.
Channel-Specific Workflow
Discord
- Use
messagewithaction="read"for each target channel before summarizing. - Prefer explicit channel IDs. If the user gives
#channel-name, includeguildId. - Read each channel separately, then merge findings into one summary.
- Do not say you need pasted messages if you can read the requested Discord channels directly.
- If the user names a Discord server but not channels, default to all concrete readable channels you can already resolve from context, tool descriptions, prior tool output, or the current server context.
- If the ask is broad and you can resolve multiple channels, prefer covering more channels over asking for narrower scope.
- If channels are explicit but timeframe is missing, read the recent bounded sample and summarize it as the latest activity instead of blocking on timeframe selection.
- Ask only if you still cannot identify any concrete readable channels without guessing.
Recommended pattern:
{"action":"read","channelId":"#announcements","guildId":"1412305846125203539","limit":50}Example user ask:
Use message read to fetch the last 50 messages from #announcements, #general, and #engineering in guild 1412305846125203539, then summarize the last 24 hours.
- Use
messagewithaction="read"for the current ingested email thread or an explicit email address. - If already in email context, omit
channelIdto read the current thread. - Email read only covers threads already ingested by the gateway.
- Do not claim mailbox-wide unread inbox access. That capability does not exist here.
Recommended patterns:
{"action":"read","limit":20}{"action":"read","channelId":"user@example.com","limit":20}- There is no first-class WhatsApp history read action yet.
- If the current session context already contains the relevant conversation, summarize that directly.
- Otherwise ask for a pasted transcript/export or clearly state that full historical WhatsApp catch-up is not available through tools yet.
- Only use
session_searchif the user wants topic-based recall and accepts approximate recall rather than a strict "last N messages" summary.
TUI / Local
- There is no first-class TUI history read action yet.
- If the needed content is already in the current session context, summarize it directly.
- Otherwise ask for a transcript/export or narrower current-session scope.
- Do not pretend you can fetch arbitrary historical TUI messages when no direct read path exists.
Synthesis Rules
After reading the scoped content:
- lead with the few most important updates
- separate facts from inferred significance
- include decisions, blockers, and follow-ups if present
- compress repetition
- do not quote large message dumps unless the user asks
Default summary shape:
- top 3-7 important updates
- decisions made
- open questions / blockers
- action items
If the user asks for "short", keep it to a tight executive summary.
Constraints
- Never summarize unseen content as if you read it.
- Do not use
session_searchas a substitute for chronological channel reads when the user explicitly asked for the last N messages from Discord or email. - For multi-channel Discord catchups, read first, summarize second.
- Prefer doing a broad useful catch-up over asking for scope. If in doubt, cover all readable channels or threads with visible activity that you can actually resolve, then mention your assumption afterward. Do not invent channels, guilds, or mailbox access that the tools did not reveal.
- For email catchups, preserve thread context and reply in normal email style if responding by email.
- If replying by email, append a polished corporate signature block derived from
IDENTITY.mddetails already loaded in context. Prefer full name, role, organization, and any real contact details that are present. Do not invent contact details, and do not use emoji or mascot-style sign-offs.
In this repository91 other files it gives its agents.
- AGENTS.mdAGENTS.md
- skills/llm-wiki/templates/AGENTS.mdAGENTS.md
- templates/AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- 1passwordSkill
- agentic-tpmSkill
- airtableSkill
- alexaSkill
- apple-calendarSkill
- apple-musicSkill
- apple-passwordsSkill
- bahnSkill
- blinkSkill
- bookingSkill
- brand-voiceSkill
- byd-batterySkill
- code-reviewSkill
- code-simplificationSkill
- current-timeSkill
- diagramSkill
- discordSkill
- distil-pii-redactorSkill
- docxSkill
- download-platform-invoicesSkill
- excalidrawSkill
- fastbillSkill
- fax-sendSkill
- feature-planningSkill
- firecrawlSkill
- froniusSkill
- ga4Skill
- gh-issuesSkill
- github-pr-workflowSkill
- gogSkill
- google-adsSkill
- google-workspaceSkill
- gwsSkill
- handoff-linksSkill
- hermes3000-writingSkill
- hetzner-cloudSkill
- hetzner-dnsSkill
- hetzner-storage-boxSkill
- heygenSkill
- homematicSkill
- hubspotSkill
- hueSkill
- human-distillSkill
- hybridclaw-helpSkill
- image-generationSkill
- inbox-cleanupSkill
- iss-positionSkill
- langfuseSkill
- lexware-officeSkill
- llm-wikiSkill
- mailchimpSkill
- manim-videoSkill
- microsoft-365Skill
- miroSkill
- mittwaldSkill
- money-savingSkill
- The first 60 shown.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
Reviews can't be read right now.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

