Reframe an AI Engagement
SkillDev toolsReframe an inherited AI delivery engagement when field evidence materially contradicts the sold brief or current boundary. Use when someone says the brief is wrong, sponsor and operator disagree, policy blocks the sold path, or a scoped human disposition is needed without rewriting history.
Available today. Use it from your connected AI after setup.
No other account needed.
Connect ahel once, and every AI you use reads what you have installed.
Then ask your AI: use the Reframe an AI Engagement skill
What this skill tells your AI
The instructions your AI receives, as published by davidahmann/applied-ai-field-guide in .agents/skills/reframe-ai-engagement/SKILL.md and read by ahel’s review.
Turn a consequential field contradiction into a bounded decision that keeps delivery moving. Do not silently rewrite the brief, infer authority, or treat this engagement disposition as production approval.
Read first
- Read Field Engagement and Accountable Reframing and the engagement-reframe schema.
- Use the engagement-reframe record, field-observation log, and discovery pack.
- Read the worked invoice engagement only as a teaching case, never as customer evidence.
- Apply
FDE-001,FDE-002,FDE-005,CTX-001,CTX-004, andDEL-002from the control catalog. - If the reframe may affect a recurring solution, resolve the solution portfolio and read only the selected context after the disposition; it is not target evidence or authority.
Workflow
- Preserve the inherited claim, exact source, revision, commercial status, and limitations without improving its wording.
- Separately verify the sponsor, process knower, operator, disposition authority, and verifier. A person may hold several roles only with separate evidence or authority bases.
- Bind the work to one engagement, customer or tenant, workflow, environment, evidence store, and retention policy. Stop on a missing or ambiguous boundary; never blend engagements.
- Inspect one actual or sanitized recent representative case and a material exception. After each interaction, stage its source record, propose cited additions, preview them for named human confirmation, then append a dated receipt that preserves rejected or deferred items.
- Classify consequential claims as
sold,stated,observed,system_enforced, orpolicy_authorized. No class universally outranks another or grants authority. - Bound the conflict: affected workflow, outcome, human authority, data, acceptance evidence, economics, downstream work, and safe behavior while unresolved.
- Prepare one cited reframe with inclusions, exclusions, retained authority, safe fallback, affected work, required decision, and next field move under each outcome.
- Obtain a scoped
continue_discovery,bounded_kickoff,defer, orstopdisposition from the verified disposition authority. Bind the authority basis, exact passage, scope, rationale, and time. - Preserve chronology and propagate only dependency-linked changes. Leave unrelated revisions and digests unchanged; route stale work for review. Derive any sponsor readout from current receipts and governed records rather than maintaining a second narrative.
Output contract
Return:
- a current field brief with roles, representative evidence, limitations, and next move;
- the competing claims, cited conflict, safe fallback, and decision brief;
- a schema-valid engagement-reframe record with scoped human disposition or explicit missing authority;
- the affected downstream list with
no_change,review_required, orsupersedeand preserved chronology; - the current boundary and evidence required to reconsider it.
- the latest interaction receipt and a source-linked current readout when the request follows a field interaction.
Stop if the relevant evidence, process knower, or disposition authority cannot be established. Do not invent observations, approvals, acceptance, or target-system authority.
Signals
- GitHub stars
- 105
- Forks
- 22
- Last commit
- Sep 2026
Advanced
- Catalog kind
- skill
- Gateway key
reframe-ai-engagement- Source
- github.com/davidahmann/applied-ai-field-guide