Ctx pack
Agent skills (open standard) for the bookends of the agent loop: context assembly, decision review, and work-shaping outputs. Works with Claude Code and any Agent Skills harness. Ships its own adoption and value measurement.
npx -y skills add arniesaha/ctxops --skill ctx-packAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
WHEN you need task-specific grounding before writing, deciding, or building. Pulls files, memory, diffs, and open threads into one pack. Triggers: 'ground in what exists', 'before I draft this', 'gather context for X', 'set up before brainstorm/decision', any moment you're tempted to dispatch parallel grounding agents. OUTPUT: a compact pack (goal, current state, decisions made, open threads, risks, next actions). PAIR BEFORE: artifact-packager, status-update, decision-review. NOT FOR: repo resume/handoff (use project-briefing), 'did we already do X' (use recall-prior-work).
SKILL.md
2.1 KB, as published. Nobody here has run it
ctx-pack
Use when grounding a specific task: drafting a doc, making a decision, planning a change. Different from project-briefing (whole-repo resume) and recall-prior-work (memory-only lookup).
Workflow
- Identify the target: project, decision, artifact, person, or workflow.
- Load the smallest useful set of sources:
- current status docs
- persistent memory for the project, if a memory tool is available (e.g. claude-mem
search/timeline; userecall-prior-workif the answer depends on prior sessions) - relevant repo files
- recent commits or diffs
- implementation notes or deviation logs from prior runs (e.g.
implementation-notes.md), if present - open tracker issues / PRs cited by the work
- deployment or runbook notes
- Separate current facts from stale assumptions.
- Produce a context pack, not a transcript dump.
Output Shape
- Goal: what this context is for.
- Current state: what is true now, with dates when useful.
- Decisions already made: avoid re-litigating them.
- Useful source files: paths or links.
- Open threads: unresolved work.
- Risks and stale assumptions: what needs verification.
- Next likely actions: 3-5 concrete options.
Rules
- Prefer status docs and curated memory over raw logs.
- Do not load private long-term memory in shared or public contexts.
- If source material conflicts, state the conflict instead of blending it.
- Date load-bearing facts. Memory is a snapshot; the repo is ground truth.
- Keep the pack brief enough to paste into another agent.