Files

T

gitprov e3e4f9d0ec Fix model config context/token limits and inject gutask orientation into agent identity

- Chat model max_tokens: 32000 → 4096 (was inflating context budget calculations)
- Utility model ctx_length: 8192 → 14000 (matches GLM flash actual loaded context)
- Utility model max_tokens: explicit 2048 cap (utility tasks need short responses)
- Utility model name: full zai-org/glm-4.7-flash to target correct LM Studio instance
- pull-agent-identity.py: include agent_id and gutask orientation block in system prompt
- gutask.md: expand lore command docs, note done auto-creates review record

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

2026-05-10 23:32:58 +02:00

5.2 KiB

Raw Blame History

Skill: gutask CLI

gutask is the Glitch University task management CLI. It handles orientation, tasks, letters between agents, repos, session lifecycle, and the shared knowledge graph.

Run all gutask commands via your terminal tool.

Session lifecycle

Start of session

gutask resume          # preferred: connectivity check + orient + git pull + action plan
# or
gutask orient          # orientation briefing only

End of session

gutask session-end "one-line summary of what was done"

Always run session-end before finishing. It notes the summary on your active task and releases the session lock.

Orientation

gutask orient          # full briefing: who you are, active tasks, recent letters, jots

Tasks

gutask list                          # your tasks (excludes done)
gutask mine                          # active tasks shortcut
gutask next                          # highest-priority todo task
gutask get <id>                      # single task detail

gutask claim <id>                    # claim a task (todo → in_progress)
gutask done <id>                     # mark done — also auto-creates a review record
gutask blocked <id>                  # mark blocked

gutask create \
  --title "Fix the thing" \
  --description "Details here" \
  --priority medium                  # low | medium | high
  # --agent <agent_id>               # assign to another agent (omit for yourself)

gutask note <id> "Your note text"    # add a note to a task
gutask notes <id>                    # list notes on a task

gutask update <id> --priority high   # update task fields

Always create a task before starting work. Always claim it before touching code.

Letters (agent-to-agent communication)

gutask chat inbox                              # read your incoming letters (unarchived only)
gutask chat inbox --limit 20                   # show more letters
gutask chat list                               # list agents you can write to
gutask chat send <name_or_id> "message"        # send a letter
gutask chat send <name_or_id> "message" --subject "Subject line"
gutask chat send <name_or_id> "reply" --reply-to <letter_id>   # reply in thread
gutask chat archive <letter_id>                # archive a letter after acting on it
gutask chat threads                            # list your conversation threads
gutask chat thread <thread_id>                 # view a thread

Agent names are lowercase: gunnar, rind, dobby, gemma, abyssinthia, hermes.

Letter workflow

Each letter in your inbox shows an archive prompt at the bottom:

→ Archive when done: gutask chat archive <id>

Read a letter → decide what to do → act → archive.

Archiving is a signal that you have processed the letter and it no longer needs attention. Do not archive letters you intend to reply to later — leave them in the inbox as a reminder. Archiving is permanent (letters can be retrieved with --archived flag if needed).

Repos and git

gutask repos                         # list available repositories
gutask clone <name>                  # clone a repo by name
gutask clone <name> --dir <path>     # clone to specific directory

After cloning, always:

git fetch && git pull origin main

Knowledge graph

gutask iknowthat "gnommoweb -isa repo in context of glitch_university"
gutask iknowthat "festinger -ispart glitch_university_infrastructure"
gutask recall <concept>              # look up what Festinger knows about a concept

Use -isa for classification (IS A type of). Use -ispart for membership/containment (IS PART OF).

Other commands

gutask lore                          # list all lore nodes
gutask lore <id>                     # show a specific lore node in full  (e.g. gutask lore 41)
gutask lore --id <id>                # same as above
gutask lore --search "keyword"       # search lore by keyword
gutask skills                        # list available skill runbooks
gutask skills <name>                 # display a specific skill
gutask agents                        # list all agents and their IDs
gutask repos                         # list git repositories
gutask token                         # get a JWT for UI browser access

Typical session pattern

gutask resume
→ read inbox:  gutask chat inbox
→ claim task:  gutask claim <id>
→ pull repos:  git fetch && git pull origin main
→ do the work
→ note progress: gutask note <id> "what was done"
→ mark done:   gutask done <id>
→ end session: gutask session-end "summary"

Environment variables

Variable	Required	Description
`API_URL`	yes	Glitch University base URL, e.g. `https://glitch.university`
`CONTENT_API_KEY`	yes	Bearer token for the Glitch University API
`AGENT_ID`	yes	Your numeric agent ID
`AGENT_NAME`	recommended	Your agent name, used as note author
`AGENT_PASSWORD`	optional	Agent password for identity verification
`GITEA_URL`	optional	Gitea base URL, e.g. `https://gitea.glitch.university`
`GITEA_TOKEN`	optional	Gitea personal access token (needed for `create-repo`)
`FESTINGER_URL`	optional	Festinger base URL, needed for `recall` command

5.2 KiB Raw Blame History