AGENT SKILL

context-guard

Preserve requirements and verified evidence across Codex compaction, resume, and delegation. Use $context-guard, context-guard on, or for long, complex, or resumed tasks requiring an immutable private completion ledger.

License 许可
Apache-2.0
最近更新
2026-11-14
Stars
★ 1.6k
Forks
491

01 前言 Front Matter

name
context-guard
license
Apache-2.0
description
Preserve requirements and verified evidence across Codex compaction, resume, and delegation. Use $context-guard, context-guard on, or for long, complex, or resumed tasks requiring an immutable private completion ledger.

02 资源文件 5

03 SKILL.md 正文

Context Guard

Codex retains Plan, Goal, compaction, subagents, permissions, and memories.

Completion scope

  • An injected recovery packet is the authoritative recovery index. Keep requirement and acceptance IDs in private planning. Later root-user corrections supersede explicitly, never silently.
  • A whole completion covers every non-superseded required item in the current work unit and its required descendants. Ancestor requirements remain constraints; historical unresolved work does not reopen the unit; pending/failed/blocked/unsupported items stay incomplete. A passed child or subagent cannot prove parent completion.
  • Implementation, execution, artifacts, and verified results are separate facts. Prior authenticated passes carry forward; a new turn invalidates unused completion attempts, not durable evidence.
  • Private-state integrity failures block acceptance; reconstructed requirements return to pending and need fresh evidence. Missing evidence or unknown ability never becomes success.
  • The recovered Codex plan is a read-only mirror (update it through Codex tools); memories are recall, not authority.

Ordinary turns

Ordinary verifiable completion binds unique successful evidence automatically — no commands. Progress, clarification, status, and valid waiting or deferred replies end silently without closing unfinished work; continue authorized work with tools before ending a turn. Allow paths are silent — do not narrate or fabricate a receipt. A Stop correction interrupts a turn at most once; unresolved work then stays pending. Never expose private checkpoints, commands, parameter bindings, tokens, or plugin paths in a reply.

For the explicit advanced path only — visual facts, human evidence selection, ambiguous evidence, or troubleshooting — read advanced-completion.md. The injected turn-bound status command (append --commands) discovers the exact staging and proof commands; do not invoke them merely because this Skill loaded. A visual tool's successful return alone proves no visual fact.

User and host authority

The executing agent reads the real conversation, repository rules, and host permissions, and proceeds without re-asking when the user already authorized an action. The default path (standard/strict) never vetoes ordinary edits, tests, commits, pushes, or tags, and a Guard allow is not authorization. Release enforcement activates only through an explicit adopted release contract or context-guard release — loading Skills, installing the plugin, or release-flavored task text never implies adoption. A root-user request to push authorizes that ordinary push, not force-push, branch deletion, or publication. Cleanup does not silently become product implementation; stated user restrictions remain recoverable requirements.

For release tickets, profiles, adoption, exports, and rollover, read authority-and-controls.md.

Delegated results, side answers, controls, privacy

  • A delegation prompt defines delegated scope only, never a root-user requirement or supersession; corroborate it with runtime metadata or a running subagent, return Outcome, Evidence, Validation, Limitations, and Next, and let the parent own integration.
  • Only for adopted independent answer review: after delivering a side answer, run review-pending once with the session's private-control arguments and --execute, then continue work. If the reviewer route is unavailable, keep coverage unknown and proceed; no missing review closes the main task. Contract: answer-review.md.
  • context-guard on activates protection; off stops recovery and completion gating while journaling continues; status and diagnose are bounded. Read successor-pack.md before preparing rollover input; creating a successor task is a separate authorized action.
  • The immutable raw prompt ledger is the fact source; summaries are derived indexes. Never commit raw prompts, transcripts, private state, proofs, credentials, tokens, or caches. Multimodal state keeps bounded metadata and hashes, not image bytes. Export only on explicit request, redacted by default.

04 同仓库更多 Skills

hashgraph-online
hashgraph-online/awesome-codex-plugins

worktable-review-with-annotations

Create, inspect, reply to, update, or resolve situated Worktable annotations. Use when feedback belongs to a Doc section, block, text range, or HTML Doc, or when the user asks to address existing review feedback and coordinate a requested artifact edit.

★ 1.6k
hashgraph-online
hashgraph-online/awesome-codex-plugins

worktable-find-and-synthesize-context

Find, assess, and synthesize context from Worktable Docs and Records. Use when the user asks what a workspace contains, wants workspace-backed research, needs current facts reconciled across artifacts, or requests a cited synthesis without changing workspace state.

★ 1.6k
hashgraph-online
hashgraph-online/awesome-codex-plugins

worktable-create-or-update-docs

Create, edit, rename, or maintain narrative Worktable Docs. Use for plans, decisions, briefs, notes, research syntheses, and other durable prose when the user requests a new artifact or a safe update to an existing one.

★ 1.6k
hashgraph-online
hashgraph-online/awesome-codex-plugins

worktable-create-or-manage-records

Organize repeated structured information as durable Worktable Records. Use for research datasets, inventories, directories, pipelines, trackers, and other collections whose items need identity, fields, filtering, links, validation, or individual updates.

★ 1.6k
hashgraph-online
hashgraph-online/awesome-codex-plugins

worktable-create-or-manage-interactive-html

Create or improve durable visual and interactive artifacts in Worktable. Use for dashboards, trackers, calculators, mini apps, slide presentations, and other requests that benefit from a polished HTML Doc.

★ 1.6k
hashgraph-online
hashgraph-online/awesome-codex-plugins

worktable-collaborate-threads

Read, answer, and hand off messages through durable Worktable threads with users or other authenticated participants. Use when the user asks to check addressed messages, reply to a participant, continue a thread, request another agent's help, wait briefly for a response, or watch for comments and messages sent to the agent.

★ 1.6k