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.

referencesrecall.md

≈693 tokens on demand. Your agent reads this file only when SKILL.md points to it.

Find / recall

Use recall when prior experience may change a consequential choice or check. Do not recall by ritual for every task. Start with the accepted intent and current source; memory cannot override either.

  1. Resolve the caller-selected project .context/ or external topic-page location and allowed owner, task, model/provider and destination. This lean route accepts public or already-cleared inputs only; it does not enforce native restricted access. Missing routing, denied access and no-match are different outcomes.
  2. For project context, start at its authored README.md topic map when present; do not create it on a read. Search metadata and relevant terms with rg in the selected bounded source. Use ordinary filesystem tools to read only likely applicable pages and the permitted support needed to judge them. Reading cleared project pages requires neither BD nor AO. Existing docs, ADRs and code remain authoritative; follow their pointers when applying a claim. CASS or native episodes are optional source locators, not an automatic raw transcript read. Follow their installed source contract when selected.
  3. Check the claim's applicability, action, support, limits and invalidation against this task's actual versions, interfaces and evidence. Confirm the exact page has the required independent support/disclosure review. Treat contradictions, withdrawn claims and unavailable support as uncertainty; do not paraphrase them into an authoritative rule.
  4. Return only the useful constraint, its support, limits and the action it changes. If none changes the next action, return no-match and continue work. Preserve indispensable acceptance when context is limited; drop optional context before it and disclose any unresolved required evidence.

The optional ao config context route supports external bundles and an explicitly bound canonical direct <consumer>/.context, retaining policy and identity bindings. It requires native BD; other consumer-overlapping roots remain refused. It is not a prerequisite for portable .context/ reading. A missing page or map does not authorize private source retrieval, an alternate store or scaffolding.

Find/recall does not capture, mine, curate, write a lesson, mutate work status or establish a benefit claim. A rare old constraint can remain valuable if its applicability and support survive. An outdated general rule may need narrow use or withdrawal, which belongs to a separately selected curation operation.

Counterexample: a page from one parser bug says an absent count was optional in that format. It cannot require all count fields to be optional in a different format; check the current contract and a discriminating example first.

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