# Digest And Explanation > Use this flagship skill to inspect ACP digest items, explanation faithfulness, routing explanations, and release readiness. - **Type:** Skill - **Install:** `agentstack add skill-jasontang-ai-acp-digest-and-explanation` - **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior - **Seller:** [jasontang-ai](https://agentstack.voostack.com/s/jasontang-ai) - **Installs:** 0 - **Category:** [Agent Skills](https://agentstack.voostack.com/c/agent-skills) - **Latest version:** 0.1.0 - **License:** Apache-2.0 - **Upstream author:** [jasontang-ai](https://github.com/jasontang-ai) - **Source:** https://github.com/jasontang-ai/acp/tree/main/skills/digest-and-explanation ## Install ```sh agentstack add skill-jasontang-ai-acp-digest-and-explanation ``` Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent. ## About # What this skill is for Use this skill when routed items must become participant-facing digests with clear, faithful, bounded explanations. # Constitutional purpose Compress routed material into participant-facing outputs without hiding disagreement, uncertainty, or source-evidence limits. # Failure modes targeted - Explanation overclaiming - False neutrality or false consensus language - Untraceable digest items - Release-ready framing unsupported by source evidence # When to use it - Reviewing intervention digests before release. - Checking `DigestItem.explanation_text` and digest-level `routing_explanations`. - Verifying that participant-facing summaries preserve disagreement and uncertainty. # When not to use it - Do not use it to select routing recipients. - Do not use it for baseline chronological thread review. - Do not rewrite canonical digest fields in skill prose. # Inputs expected - Routed intervention cycle or participant view - Digest items and routing explanations - Cycle config and release target - Optional participant id for focused review # Required evidence for valid use - Digest items with source contribution references - Digest-level and item-level explanation text - Release context sufficient to know whether the work is review-only or release-authorized - Critic or routing notes when omission/fairness concerns already exist # Preflight / prerequisites 1. Confirm routing is complete. 2. Confirm each digest item has a source contribution reference. 3. Confirm whether the work is review-only or release-authorized. 4. Load protocol contract sections for Digest and DigestItem field names. # Workflow 1. Inspect every digest and item. 2. Confirm item explanations are non-empty and traceable to routing factors. 3. Check digest-level explanations for overclaiming or consensus language. 4. Confirm bridge items are visibly bounded and justified. 5. Recommend release, revise, or escalate. # Decision rules / judgment criteria - Explanations must be faithful to actual routing factors. - Digest compression must preserve unresolved questions. - Do not hide omitted-but-relevant items when critic evidence flags them. - Participant reading burden remains a hard constraint. # Escalation rules Escalate when explanation text is missing, source references are ambiguous, bridge exposure is unexplained, or the digest implies consensus that the source material does not support. # Available scripts - `scripts/inspect-digest.mjs`: inspect digests and normalize item explanations to protocol-facing review fields. # Produced artifacts - Digest quality assessment - Explanation coverage summary - Release/revise/escalate recommendation - Participant-facing risk notes # Allowed claims - That explanation text is faithful or unfaithful to routing evidence - That a digest preserves or suppresses unresolved disagreement - That release should be blocked until explanation or source-trace repair occurs # Forbidden claims - That digest language proves consensus - That explanation polish substitutes for source support - That omitted concerns are irrelevant without evidence - That explanation faithfulness alone establishes fairness or legitimacy # Human override points - Manual digest revision before participant release - Source-reference challenge when explanation text is under-supported - Release hold when compression would misstate disagreement # Outputs - Digest quality assessment - Explanation coverage summary - Release/revise/escalate recommendation - Participant-facing risk notes # Failure handling If routing is incomplete, return to `epistemic-routing`. If explanation fields are missing, block release and request digest repair. # Trust / safety notes Digest explanations affect perceived legitimacy. Treat vague or overconfident explanations as governance risks. # Composition notes Use after `epistemic-routing` and before `operator-audit`. It is a flagship dependency for public-hearing and Foresight workflows. # Examples to inspect next Read `examples/casebook.md` for flagship digest review cases, including faithful explanation and false-consensus anti-patterns. # Evaluation hooks Run `scripts/inspect-digest.mjs` and `npm run skills:audit`. Use `evals/rubric.md` for explanation faithfulness review. ## Source & license This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code. - **Author:** [jasontang-ai](https://github.com/jasontang-ai) - **Source:** [jasontang-ai/acp](https://github.com/jasontang-ai/acp) - **License:** Apache-2.0 Install and usage instructions live in the source repository linked above. ## Pricing - **Free** — Free ## Security capabilities Automated source analysis of v0.1.0 — what this tool can access: - **Network access:** no - **Filesystem access:** no - **Shell / process execution:** no - **Environment & secrets:** no - **Dynamic code execution:** no *"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.* ## Versions - **0.1.0** — security scan: passed — Imported from the upstream source. ## Links - Listing page: https://agentstack.voostack.com/l/skill-jasontang-ai-acp-digest-and-explanation - Seller: https://agentstack.voostack.com/s/jasontang-ai - Browse the marketplace: https://agentstack.voostack.com/browse --- Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.