docs(skill): match everyday Life requests and document deletion safely #119

New issue

Closed

opened 2026-09-23 23:45:28 +00:00 by yuris · 0 comments

yuris commented

2026-09-23 23:45:28 +00:00

Owner

Why

The skill description currently mentions commitments, planning, interruptions, completion, and rationale, but omits common phrases such as calendar, schedule, open tasks, and delete an event. An agent may not select the skill for ordinary Life requests. The current skill examples also omit the required client config path, although the installed CLI does not auto-load a token or config; calling life tasks list without --config fails. When explicit deletion exists, agents must distinguish it from cancellation and obtain the live ETag before a destructive action.

Scope

Expand skills/life/SKILL.md frontmatter in plain language for calendar, agenda, schedule, task/todo, free time, planning, next action, completion, cancellation, deletion, and contextual why requests. Keep the description specific to personal Life data to avoid matching general calendar programming questions.
State that each CLI invocation needs a configured client path or an equivalent deployment-provided invocation that loads the scoped token; do not embed this user's paths, hosts, or credentials in the portable skill. Correct references/commands.md if it implies a default config that the binary does not provide.
After the delete commands land, document their CLI syntax, write scope, difference from cancel, quoted ETag, stable operation ID, recurrence-series safeguards, and explicit user approval. Continue to treat ambiguous references and partial outcomes conservatively; link to the existing error/recovery reference.
Incorporate the bounded, exit-code-safe OMP presentation guidance from the separate preview issue without claiming the skill can suppress harness-rendered tool cards. Preserve the normal JSON envelope contract and cross-agent portability.

Acceptance

Try representative prompts for today's agenda, open tasks, deleting a calendar event, and completing a task; verify the skill is selected and the correct read/approval behavior is followed. Do not claim a prompt-routing test if the harness cannot exercise it.
Existing skill package checks pass and every documented delete command is present in the real Clap tree. No host-specific topology, secret values, or speculative OMP settings enter the package.

Dependency

Complete the deletion and preview issues before the final skill text is merged, since they determine what commands and presentation behavior it can honestly document.

## Why The skill description currently mentions commitments, planning, interruptions, completion, and rationale, but omits common phrases such as calendar, schedule, open tasks, and delete an event. An agent may not select the skill for ordinary Life requests. The current skill examples also omit the required client config path, although the installed CLI does not auto-load a token or config; calling life tasks list without --config fails. When explicit deletion exists, agents must distinguish it from cancellation and obtain the live ETag before a destructive action. ## Scope - Expand skills/life/SKILL.md frontmatter in plain language for calendar, agenda, schedule, task/todo, free time, planning, next action, completion, cancellation, deletion, and contextual why requests. Keep the description specific to personal Life data to avoid matching general calendar programming questions. - State that each CLI invocation needs a configured client path or an equivalent deployment-provided invocation that loads the scoped token; do not embed this user's paths, hosts, or credentials in the portable skill. Correct references/commands.md if it implies a default config that the binary does not provide. - After the delete commands land, document their CLI syntax, write scope, difference from cancel, quoted ETag, stable operation ID, recurrence-series safeguards, and explicit user approval. Continue to treat ambiguous references and partial outcomes conservatively; link to the existing error/recovery reference. - Incorporate the bounded, exit-code-safe OMP presentation guidance from the separate preview issue without claiming the skill can suppress harness-rendered tool cards. Preserve the normal JSON envelope contract and cross-agent portability. ## Acceptance - Try representative prompts for today's agenda, open tasks, deleting a calendar event, and completing a task; verify the skill is selected and the correct read/approval behavior is followed. Do not claim a prompt-routing test if the harness cannot exercise it. - Existing skill package checks pass and every documented delete command is present in the real Clap tree. No host-specific topology, secret values, or speculative OMP settings enter the package. ## Dependency Complete the deletion and preview issues before the final skill text is merged, since they determine what commands and presentation behavior it can honestly document.

yuris referenced this issue from a pull request that will close it,

2026-09-24 07:49:18 +00:00

docs(skill): expand discovery triggers, clean output guidance, and deletion contract #122

yuris closed this issue

2026-09-24 08:42:34 +00:00

yuris referenced this issue from a commit

2026-09-24 08:42:35 +00:00

docs(skill): expand discovery triggers, clean output guidance, and deletion contract (#122)