outcome-reporting
Report task outcomes after using recall/search results so engram can measure whether memory helps. Trigger after completing a task that used recall, search, or symbol_context results.
SKILL.md.agents/skills/outcome-reporting/SKILL.md
GitHubOutcome Reporting Skill
When you use engram's recall, search, or symbol_context to find information
and then complete a task based on that information, report the outcome.
When to report
After completing ANY task where you used engram retrieval results:
- Found the right file/symbol → report success
- Found partial information → report success with note
- Found nothing useful → report failure
- Found WRONG information → report failure (interference data!)
How to report
report_outcome({
"query_id": "<from the recall response header>",
"success": true|false,
"capability": "<what task you were doing>",
"recalled_targets": ["<symbols/files that helped>"],
"verifier": "<your agent name>"
})
Why this matters
Every outcome feeds the ΔP measurement — the proof that memory improves agent performance. Report honestly: failures are MORE valuable than successes because they identify interference and gaps.
Check the aggregate
Run outcome_report periodically to see success rates by capability
and interference candidates.