All skills
boshu2 avatar

/memory

@92e6f5d
by Boboshu2/agentops446 stars
42

Find reviewed context, capture evidence or curate maintained claims. Use when: prior evidence can change an action, or learning is requested; no mandatory recall or lesson.

Use this Skill: https://skilld.dev/gh/boshu2/agentops/memory

This session only. Nothing lands on disk.

referencescurate.md

≈885 tokens on demand. Your agent reads this file only when SKILL.md points to it.

Curate / qualify / retire

Maintain caller-selected reviewed Markdown topic pages in project .context/ or an external bundle. BD remains the work/status/handoff owner, Git the content-history owner, and native/CASS systems the episode owner. Existing docs, ADRs and code retain their declared authority. Link those owners; do not duplicate policy, build a second tracker or copy raw history. A small authored README.md topic map is a navigation aid, not a work/status index; keep it only as useful to the selected pages. This lean operation accepts public or already-cleared inputs only; it supplies no native isolation for restricted sources.

  1. Find the existing topic page and inspect its current claims and review evidence before editing. Reuse/update it instead of a lesson-per-session file. If no relevant page exists and the caller selected a destination, create one topic page only for a concrete reusable claim and consumer. Do not scaffold empty directories, import private material or create a store automatically. Ordinary filesystem reads of cleared project pages need neither BD nor AO; native source operations retain their own requirements.
  2. Draft the smallest change in protected external non-Git staging. Each entry gives applicability, action, support, limits and invalidation. Cite exact source identities sufficient to inspect evidence without copying private material. Keep unrelated claims and useful rare constraints.
  3. Qualify a single incident narrowly. General methods require stronger support and later reapplication evidence; contradiction may narrow or remove a rule. Retire when invalidated or unsupported after examination, not merely old. Preserve withdrawal reason, provenance and unique evidence in the existing page/history under owner policy. No blind TTL or deletion sweep.
  4. Have a fresh authorized author-distinct context review exact factual support and exact destination disclosure before any Git object, index, stash or import. Include intended paths and metadata in the reviewed payload. A changed claim or destination requires matching review; the author cannot approve their own knowledge. Public input does not waive factual review.
  5. Apply only the approved content to the caller-selected destination under existing Git authority. Read back exact bytes and confirm that citations and withdrawal facts remain available. Do not commit/push unless authorized. If review, routing or permission is missing, return the supported gap with the protected draft; do not invent an alternate memory destination. Project placement does not move drafts or review proof into the checkout. The optional ao config context route supports an explicitly bound canonical direct <consumer>/.context or an external bundle; it requires native BD and retains policy and identity bindings. A resolved route does not approve page admission.

Ordinary Markdown is sufficient. If the caller selects the existing OKF profile, use its profile and ao provenance check-okf --file <topic.md> for structure. This checks no factual support, disclosure, isolation, review or usefulness; the profile is optional.

Return the changed claim and why, its limits and exact review evidence, or no-change. Curation can remove rules. It cannot alter earlier product verdicts or claim benefit until later task evidence shows helpful reuse. A page admitted for one owner/destination is not authorized for another.

Source: SKILL.md on GitHub

No alerts8d3 checks · Risk SAFE
  • Gen Agent Trust Hub8d

    The skill facilitates memory retrieval and learning by reading and writing external topic pages. This design introduces an indirect prompt injection surface, which the skill attempts to mitigate through mandatory review processes. It also includes references to a log-mining script.

  • Socket8d

    No alerts

  • Snyk8d

    Risk: LOW · No issues

Signed by skilld at 92e6f5d. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub yesterday.

Activeupdated last week

README badge

README badge for boshu2/agentops/memory